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,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.01,0.19211584091186523
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.01,0.28341312408447267
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.01,0.3351916885375977
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.01,0.47300479888916014
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.01,0.5995775985717773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.01,0.6050444793701172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.01,0.5934643173217774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.01,0.6003993606567383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.01,0.6107046508789062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.01,0.6024684906005859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.01,0.6239020919799805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.01,0.6344448089599609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.01,0.6721548461914063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.01,0.7242867279052734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.01,0.7745113372802734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.01,0.9312326049804687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.01,1.0838956451416015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.01,1.3706854248046876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.01,1.6811219787597658
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.01,2.315856628417969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.01,2.9573043823242187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.01,4.300771179199218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.01,5.756649169921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.01,9.602637939453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.01,13.956832275390624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.01,22.62718505859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.01,32.6762060546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.01,0.11821375846862794
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.01,0.16349504470825196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.01,0.18515520095825194
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.01,0.26298175811767577
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.01,0.2784723281860352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.01,0.27908863067626954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.01,0.2810758399963379
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.01,0.279116153717041
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.01,0.2861555290222168
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.01,0.28815359115600586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.01,0.2882643127441406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.01,0.2807923126220703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.01,0.29700288772583006
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.01,0.33339519500732423
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.01,0.3465260696411133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.01,0.353004150390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.01,0.3778598403930664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.01,0.44113342285156254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.01,0.5329190444946289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.01,0.6937388610839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.01,0.8336045074462891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.01,1.1490156555175781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,1,1,power_law_1.01,42.097065429687504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.01,1.487407989501953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.01,2.189954528808594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.01,2.799896850585937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.01,4.152430114746094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.01,5.44354736328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,1,1,power_law_1.01,6.94251220703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.01,0.10530688285827636
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.01,0.14973567962646483
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.01,0.1495852756500244
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.01,0.15072704315185548
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.01,0.14904704093933105
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.01,0.05444479942321777
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.01,0.14926912307739257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.01,0.08110591888427734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.01,0.15262271881103515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.01,0.15382911682128905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.01,0.14704064369201658
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.01,0.15386495590209961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.01,0.17696960449218752
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.01,0.1607302474975586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.01,0.18533119201660156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.01,0.19410367965698244
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.01,0.21486272811889648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.01,0.27162496566772465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.01,0.3202278518676758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.01,0.42102401733398437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.01,0.5127718353271484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.01,0.71272705078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.01,0.9230329895019531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.01,1.3266802978515624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.01,1.7256838989257812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,1,1,power_law_1.2,0.1985286331176758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,1,1,power_law_1.2,0.28285888671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,1,1,power_law_1.2,0.3048236846923828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,1,1,power_law_1.2,0.4736684799194336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,1,1,power_law_1.2,0.5766636657714843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,1,1,power_law_1.2,0.6047993469238281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,1,1,power_law_1.2,0.6057798385620117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,1,1,power_law_1.2,0.5963206481933594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,1,1,power_law_1.2,0.6103270339965821
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,1,1,power_law_1.2,0.6117043304443359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,1,1,power_law_1.2,0.6372358322143554
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,1,1,power_law_1.2,0.6534413146972656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,1,1,power_law_1.2,0.6586246490478516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,1,1,power_law_1.2,0.7334233856201171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,1,1,power_law_1.2,0.7768851470947266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,1,1,power_law_1.2,0.97159423828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,1,1,power_law_1.2,1.0845394897460938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.01,2.562622680664062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,1,1,power_law_1.2,1.379156494140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,1,1,power_law_1.2,1.6883564758300782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,1,1,power_law_1.2,2.309520568847656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,1,1,power_law_1.2,2.9525665283203124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,1,1,power_law_1.2,4.299112854003907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,1,1,power_law_1.2,5.87244873046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.01,3.427736206054688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,1,1,power_law_1.2,9.980621337890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,1,1,power_law_1.2,14.08319580078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.01,4.22767822265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,1,1,power_law_1.2,0.11570879936218262
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,1,1,power_law_1.2,0.15716095924377443
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,1,1,power_law_1.2,0.16971008300781248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,1,1,power_law_1.2,0.2606540870666504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,1,1,power_law_1.2,0.26673408508300783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,1,1,power_law_1.2,0.27693567276000974
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,1,1,power_law_1.2,0.2794278335571289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,1,1,power_law_1.2,0.27683456420898434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,1,1,power_law_1.2,0.2824831962585449
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,1,1,power_law_1.2,0.28448064804077144
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,1,1,power_law_1.2,0.2872998428344727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,1,1,power_law_1.2,0.2900044822692871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,1,1,power_law_1.2,0.29236223220825197
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,1,1,power_law_1.2,0.33236801147460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,1,1,power_law_1.2,0.35606849670410157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,1,1,power_law_1.2,0.36703296661376955
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,1,1,power_law_1.2,0.38171264648437503
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,1,1,power_law_1.2,0.4389779281616211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,1,1,power_law_1.2,0.5316460800170899
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,1,1,power_law_1.2,0.6653292846679688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,1,1,power_law_1.2,0.8249075317382812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,1,1,power_law_1.2,22.8561083984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,1,1,power_law_1.2,1.1524864196777345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,1,1,power_law_1.2,1.5171098327636718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,1,1,power_law_1.2,2.1591539001464843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,1,1,power_law_1.2,2.8719207763671877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,1,1,power_law_1.2,32.872089843750004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,1,1,power_law_1.2,4.159703674316406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,1,1,power_law_1.2,0.05455679893493652
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,1,1,power_law_1.2,0.07794303894042968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,1,1,power_law_1.2,0.08915264129638671
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,1,1,power_law_1.2,0.14745856285095216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,1,1,power_law_1.2,0.14540224075317382
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,1,1,power_law_1.2,0.15040127754211424
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,1,1,power_law_1.2,0.1514515209197998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,1,1,power_law_1.2,0.1489523220062256
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,1,1,power_law_1.2,0.15255552291870117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,1,1,power_law_1.2,0.1526655960083008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,1,1,power_law_1.2,0.15445823669433595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,1,1,power_law_1.2,0.15603648185729982
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,1,1,power_law_1.2,0.1614905548095703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,1,1,power_law_1.2,0.17783231735229493
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,1,1,power_law_1.2,0.18949951171874999
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,1,1,power_law_1.2,0.20232320785522462
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,1,1,power_law_1.2,41.6021435546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,1,1,power_law_1.2,0.2743123245239258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,1,1,power_law_1.2,5.578555297851563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,1,1,power_law_1.2,0.2205753517150879
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,1,1,power_law_1.2,6.841801147460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,1,1,power_law_1.2,0.321952018737793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,1,1,power_law_1.2,0.42312255859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,1,1,power_law_1.2,0.5205535888671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,1,1,power_law_1.2,0.7289446258544923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,1,1,power_law_1.2,0.9246431732177735
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,1,1,power_law_1.2,1.3662176513671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,1,1,power_law_1.2,1.7669279479980466
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.01,0.30878143310546874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.01,0.45520896911621095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.01,0.5246451187133789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.01,0.8009798431396484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.01,0.9077241516113281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.01,0.9142752075195313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.01,0.9199501037597656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.01,0.9151404571533203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.01,0.9265542602539062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.01,0.9557683563232422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.01,1.0345318603515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.01,1.1178399658203124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.01,1.229580764770508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.01,1.3131455993652343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.01,1.3873663330078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.01,1.4949005126953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.01,1.8038662719726564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.01,2.2852934265136717
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.01,2.8198693847656253
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,1,1,power_law_1.2,2.60015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.01,3.9272518920898434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.01,5.03873291015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.01,7.405499877929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.01,10.24223388671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,1,1,power_law_1.2,3.434407043457031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.01,16.706300048828126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.01,24.26458984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.01,39.3232421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.01,0.157323522567749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.01,0.23749568939208981
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.01,0.2831699180603027
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.01,0.4327251052856445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.01,0.46016830444335943
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,1,1,power_law_1.2,4.333216552734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.01,0.463061752319336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.01,0.46813377380371096
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.01,0.4624127960205078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.01,0.46352256774902345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.01,0.4733369445800781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.01,0.47700992584228513
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.01,0.47158336639404297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.01,0.4859283065795898
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.01,0.5018848037719726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.01,0.5077689743041992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.01,0.524777603149414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.01,0.572481918334961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.01,0.6850835418701171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.01,0.8028787231445312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.01,1.0397471618652343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.01,1.2617708587646486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.01,1.8388038635253907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.01,2.2791072082519532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.01,3.4738330078125004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.01,57.34404296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.01,4.3585031127929685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.01,6.677184448242187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.01,8.613823852539062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,1,1,power_law_1.01,73.998896484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,1,1,power_law_1.01,10.903472900390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.01,0.08003392219543456
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.01,0.24522943496704103
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.01,0.24539903640747068
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.01,0.24669824600219728
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.01,0.2427155113220215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.01,0.24324031829833986
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.01,0.24879039764404295
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.01,0.2503910446166992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.01,0.13084287643432618
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.01,0.15661952018737793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.01,0.24014656066894532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.01,0.24713216781616212
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.01,0.25991167068481447
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.01,0.29173311233520505
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.01,0.30012800216674806
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.01,0.31698175430297854
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.01,0.3579276657104492
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.01,0.45708545684814456
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.01,0.5330400085449218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.01,0.6977503967285157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.01,0.8608704376220704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.01,1.179296646118164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.01,1.502030792236328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.01,2.213976287841797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.01,2.8346981811523437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,1,1,power_law_1.2,0.30763391494750975
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,1,1,power_law_1.2,0.43601921081542966
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,1,1,power_law_1.2,0.4882854461669922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,1,1,power_law_1.2,0.8024915313720703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,1,1,power_law_1.2,0.8801773071289063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,1,1,power_law_1.2,0.910561294555664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,1,1,power_law_1.2,0.9170559692382814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,1,1,power_law_1.2,0.9112191772460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,1,1,power_law_1.2,0.935186538696289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,1,1,power_law_1.2,0.950140151977539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,1,1,power_law_1.2,1.0380204772949218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,1,1,power_law_1.2,1.1607437133789062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,1,1,power_law_1.2,1.2109747314453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,1,1,power_law_1.2,1.3079916381835939
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,1,1,power_law_1.2,1.3761996459960937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,1,1,power_law_1.2,1.5050265502929687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,1,1,power_law_1.2,1.8064486694335937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.01,4.123138427734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,1,1,power_law_1.2,2.292790985107422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,1,1,power_law_1.2,2.8217752075195315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,1,1,power_law_1.2,3.904526062011719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,1,1,power_law_1.2,5.067717895507813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.01,5.3644433593750005
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,1,1,power_law_1.2,7.4434356689453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,1,1,power_law_1.2,10.576304931640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,1,1,power_law_1.2,17.161917724609374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.01,6.727825927734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,1,1,power_law_1.2,24.89694580078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,1,1,power_law_1.2,0.15737279891967773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,1,1,power_law_1.2,0.2298393630981445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,1,1,power_law_1.2,0.25632064819335937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,1,1,power_law_1.2,0.43257152557373046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,1,1,power_law_1.2,0.4467449569702149
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,1,1,power_law_1.2,0.4632070541381836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,1,1,power_law_1.2,0.4672198486328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,1,1,power_law_1.2,0.45957504272460936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,1,1,power_law_1.2,0.4714937591552735
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,1,1,power_law_1.2,0.4707072067260742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,1,1,power_law_1.2,0.4762086486816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,1,1,power_law_1.2,0.4788991928100586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,1,1,power_law_1.2,0.47961982727050784
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,1,1,power_law_1.2,0.5018598556518554
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,1,1,power_law_1.2,0.515412483215332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,1,1,power_law_1.2,0.5372025680541992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,1,1,power_law_1.2,0.5917254257202148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,1,1,power_law_1.2,0.6801318359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,1,1,power_law_1.2,0.7975577545166016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,1,1,power_law_1.2,1.0186534118652344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,1,1,power_law_1.2,1.2718323516845704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,1,1,power_law_1.2,1.7976985168457031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,1,1,power_law_1.2,39.630771484374996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,1,1,power_law_1.2,2.278542022705078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,1,1,power_law_1.2,3.3496084594726563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,1,1,power_law_1.2,4.3163320922851565
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,1,1,power_law_1.2,57.14951171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,1,1,power_law_1.2,6.455855102539063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,1,1,power_law_1.2,0.08370944023132323
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,1,1,power_law_1.2,0.12493887901306153
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,1,1,power_law_1.2,0.142360315322876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,1,1,power_law_1.2,0.23966527938842774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,1,1,power_law_1.2,0.23538560867309571
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,1,1,power_law_1.2,0.24489471435546878
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,1,1,power_law_1.2,0.2461427116394043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,1,1,power_law_1.2,0.24201663970947265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,1,1,power_law_1.2,0.24725055694580078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,1,1,power_law_1.2,0.2483564758300781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,1,1,power_law_1.2,0.24988096237182617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,1,1,power_law_1.2,0.25225599288940426
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,1,1,power_law_1.2,0.26438207626342775
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,1,1,power_law_1.2,0.29365375518798825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,1,1,power_law_1.2,8.648740234375001
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,1,1,power_law_1.2,76.648837890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,1,1,power_law_1.2,10.804371337890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,1,1,power_law_1.2,0.31007551193237304
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,1,1,power_law_1.2,0.32879295349121096
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,1,1,power_law_1.2,0.3700070571899414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,1,1,power_law_1.2,0.4605305480957031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,1,1,power_law_1.2,0.5359673690795899
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,1,1,power_law_1.2,0.6969261169433594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,1,1,power_law_1.2,0.8661126708984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,1,1,power_law_1.2,1.2057350158691407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,1,1,power_law_1.2,1.5074412536621093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,1,1,power_law_1.2,2.19706298828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,1,1,power_law_1.2,2.898700866699219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.01,0.1423731231689453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.01,0.2794668769836426
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.01,0.49113086700439457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.01,0.9011615753173828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.01,1.3524479675292969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.01,1.7396185302734373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.01,2.4691647338867186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.01,2.6074508666992187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.01,2.6650592041015626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.01,2.6961575317382813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.01,2.8435699462890627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.01,2.989966735839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,1,1,power_law_1.2,4.158084411621094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.01,3.0664691162109374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.01,3.2592614746093753
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.01,3.3536608886718753
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.01,3.5631057739257814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.01,3.387943725585937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.01,3.835618591308594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.01,4.129710083007812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.01,4.680477294921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.01,5.212191772460938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.01,6.370881347656249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.01,7.612229003906251
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,1,1,power_law_1.2,5.478275146484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.01,10.31416748046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.01,12.753680419921874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.01,17.8250244140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.01,0.09232704162597656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.01,23.2362548828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.01,0.26588352203369137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,1,1,power_law_1.01,28.22867431640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.01,0.4778195190429687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.01,0.1500825595855713
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.01,0.9340525054931641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.01,1.351432342529297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.01,1.2831922912597657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,1,1,power_law_1.2,6.73882568359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.01,1.3797132873535156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.01,1.3975552368164064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.01,1.5054220581054687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.01,0.7090112304687499
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.01,1.523120574951172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.01,1.5402342224121095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.01,1.460064697265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.01,1.6806207275390626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.01,1.6744345092773436
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.01,1.79176513671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.01,1.737953338623047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.01,1.883783721923828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.01,2.0160850524902343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.01,2.1457395935058594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.01,2.4370156860351564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.01,2.763823852539063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.01,3.4551168823242184
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.01,4.176617736816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.01,5.56935302734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.01,6.891707763671874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,1,1,power_law_1.01,8.311340942382813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.01,0.08696640014648438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.01,0.43475006103515623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.01,0.12116607666015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.01,0.5559743881225586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.01,0.17495359420776366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.01,0.7881037139892578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.01,0.832995834350586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.01,0.8614604949951172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.01,0.8517638397216796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.01,0.29622592926025393
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.01,0.9052294158935548
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.01,0.9331648254394531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.01,0.9463468933105469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.01,0.9629151916503906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.01,0.9959398651123047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.01,1.0052928161621093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.01,1.0533932495117189
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.01,1.1207078552246093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.01,1.1967955017089844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.01,1.3349766540527344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.01,1.4840562438964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.01,1.77085693359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.01,2.1231372070312498
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.01,2.8271691894531252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.01,3.586355895996094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,1,1,power_law_1.2,0.1393139171600342
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,1,1,power_law_1.2,0.290885124206543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,1,1,power_law_1.2,0.5086751937866211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,1,1,power_law_1.2,0.8372019195556641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,1,1,power_law_1.2,1.239504623413086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,1,1,power_law_1.2,1.6424345397949218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,1,1,power_law_1.2,2.422363586425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,1,1,power_law_1.2,2.551580810546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,1,1,power_law_1.2,2.617662658691406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,1,1,power_law_1.01,5.112389831542968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,1,1,power_law_1.2,2.715203857421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,1,1,power_law_1.01,6.601044921874999
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,1,1,power_law_1.2,2.8329901123046874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,1,1,power_law_1.01,8.084019165039063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,1,1,power_law_1.2,2.9728472900390623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,1,1,power_law_1.2,3.1142791748046874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,1,1,power_law_1.2,3.401820068359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,1,1,power_law_1.2,3.4381011962890624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,1,1,power_law_1.2,3.5085504150390627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,1,1,power_law_1.2,3.6505728149414063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,1,1,power_law_1.2,3.98498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,1,1,power_law_1.2,4.266752624511719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,1,1,power_law_1.2,4.873484191894532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,1,1,power_law_1.2,5.371977539062501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,1,1,power_law_1.2,6.590205688476563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,1,1,power_law_1.2,7.910491333007812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,1,1,power_law_1.2,10.37741455078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,1,1,power_law_1.2,13.119827880859376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,1,1,power_law_1.2,0.09140735626220703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,1,1,power_law_1.2,0.14963711738586424
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,1,1,power_law_1.2,0.265930233001709
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,1,1,power_law_1.2,0.4430796813964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,1,1,power_law_1.2,0.6506771087646485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,1,1,power_law_1.2,18.207049560546874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,1,1,power_law_1.2,0.8826719665527344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,1,1,power_law_1.2,1.2592301177978515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,1,1,power_law_1.2,1.3242124938964843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,1,1,power_law_1.2,1.3556568908691407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,1,1,power_law_1.2,1.4038450622558594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,1,1,power_law_1.2,1.4425651550292968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,1,1,power_law_1.2,23.4903369140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,1,1,power_law_1.2,1.486965789794922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,1,1,power_law_1.2,1.5227027893066407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,1,1,power_law_1.2,1.5721273803710938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,1,1,power_law_1.2,1.6946566772460936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,1,1,power_law_1.2,1.7336268615722656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,1,1,power_law_1.2,1.7650355529785156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,1,1,power_law_1.2,1.8775462341308593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,1,1,power_law_1.2,1.9368389892578126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,1,1,power_law_1.2,2.102902374267578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,1,1,power_law_1.2,2.239554595947266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,1,1,power_law_1.2,28.822900390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,1,1,power_law_1.2,2.5473260498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,1,1,power_law_1.2,2.8885919189453126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,1,1,power_law_1.2,3.560373229980469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,1,1,power_law_1.2,4.276004333496094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,1,1,power_law_1.2,0.08704575538635254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,1,1,power_law_1.2,5.660645751953124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,1,1,power_law_1.2,0.12021056175231934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,1,1,power_law_1.2,0.17216127395629882
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,1,1,power_law_1.2,7.0195507812499995
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,1,1,power_law_1.2,0.27370687484741213
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,1,1,power_law_1.2,0.3959910583496094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,1,1,power_law_1.2,0.5218828964233398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,1,1,power_law_1.2,0.7729325103759765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,1,1,power_law_1.2,0.8134086608886719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,1,1,power_law_1.2,8.5669970703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,1,1,power_law_1.2,0.8382514953613281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,1,1,power_law_1.2,0.868062744140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,1,1,power_law_1.2,0.896903076171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,1,1,power_law_1.2,0.9287801361083984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,1,1,power_law_1.2,0.9521510314941406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,1,1,power_law_1.2,0.9886758422851563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,1,1,power_law_1.2,1.0220819091796876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,1,1,power_law_1.2,1.0536224365234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,1,1,power_law_1.2,1.0987725067138672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,1,1,power_law_1.2,1.2159500885009766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,1,1,power_law_1.2,1.3179257202148438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,1,1,power_law_1.2,1.5471455383300783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,1,1,power_law_1.2,1.7522668457031252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,1,1,power_law_1.2,2.248865966796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,1,1,power_law_1.2,2.8034689331054685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,1,1,power_law_1.2,3.905447082519531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.01,0.0514470386505127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.01,0.0767302417755127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.01,0.10969599723815918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.01,0.12955583572387694
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.01,0.1866547203063965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.01,0.19535039901733398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.01,0.1989286422729492
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.01,0.21111167907714842
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.01,0.2145267105102539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.01,0.22379135131835937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.01,0.23648191452026368
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.01,0.23956991195678712
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.01,0.24633344650268554
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.01,0.2621824073791504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.01,0.26809856414794925
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.01,0.27878143310546877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.01,0.3039187240600586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.01,0.33417472839355467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.01,0.40329025268554686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.01,0.48092929840087895
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.01,0.6380006408691405
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.01,0.7955693054199219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.01,1.1183071899414063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.01,1.442611846923828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.01,2.091716461181641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,1,1,power_law_1.2,5.099244689941406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.01,2.739840087890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,1,1,power_law_1.01,3.3808563232421873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.01,0.03286144018173218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.01,0.06276991844177246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.01,0.07834496021270751
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.01,0.09545856475830078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.01,0.13183744430541994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.01,0.13754688262939455
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.01,0.13932736396789552
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.01,0.14300736427307129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.01,0.1425811195373535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.01,0.0400710391998291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.01,0.14611328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.01,0.1643507194519043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.01,0.16565824508666993
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.01,0.16798208236694337
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.01,0.17433727264404297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.01,0.17681087493896483
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.01,0.18774400711059572
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.01,0.1940287971496582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.01,0.21018751144409178
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.01,0.24036224365234374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.01,0.2812236785888672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.01,0.3485728073120117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.01,0.42104129791259765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.01,0.5668844985961914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.01,0.7179551696777343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.01,1.0123340606689453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.01,1.3092710876464844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,1,1,power_law_1.01,1.595203857421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.01,0.03549567937850952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.01,0.046672639846801756
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.01,0.04076032161712646
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.01,0.05366399765014649
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.01,0.05995967864990235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.01,0.07791232109069825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.01,0.0821446418762207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.01,0.08378815650939941
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.01,0.0857862377166748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.01,0.0868012809753418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.01,0.08988608360290527
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.01,0.09164544105529784
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.01,0.09308287620544434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.01,0.09714752197265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.01,0.1013651180267334
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.01,0.10436863899230957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.01,0.11079936027526856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.01,0.12584256172180175
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.01,0.14551424026489257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.01,0.1830335998535156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.01,0.22113407135009763
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.01,0.3022630310058594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.01,0.5680300903320312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.01,0.38900096893310543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.01,0.7458048248291016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.01,1.116970901489258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,1,1,power_law_1.2,0.0356710410118103
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,1,1,power_law_1.2,0.05194687843322754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,1,1,power_law_1.2,0.07381887912750244
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,1,1,power_law_1.2,0.1261689567565918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.01,1.5040179443359374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,1,1,power_law_1.2,0.1032953643798828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,1,1,power_law_1.2,0.19811519622802734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,1,1,power_law_1.2,0.19035711288452148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,1,1,power_law_1.2,0.21549055099487308
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,1,1,power_law_1.2,0.21572544097900392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,1,1,power_law_1.2,0.22639232635498047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,1,1,power_law_1.01,1.88586181640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,1,1,power_law_1.2,0.24336767196655273
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,1,1,power_law_1.2,0.26720256805419923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,1,1,power_law_1.2,0.25590463638305666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,1,1,power_law_1.2,0.27113407135009765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,1,1,power_law_1.2,0.2061574363708496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,1,1,power_law_1.2,0.27969343185424805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,1,1,power_law_1.2,0.23849855422973634
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,1,1,power_law_1.2,0.33927680969238283
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,1,1,power_law_1.2,0.4136921691894531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,1,1,power_law_1.2,0.4839091110229492
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,1,1,power_law_1.2,0.6430521392822266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,1,1,power_law_1.2,0.31324287414550783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,1,1,power_law_1.2,0.8075718688964845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,1,1,power_law_1.2,1.1261491394042968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,1,1,power_law_1.2,1.4500186157226562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,1,1,power_law_1.2,0.03288383960723877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,1,1,power_law_1.2,0.04108799934387207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,1,1,power_law_1.2,0.06270207881927489
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,1,1,power_law_1.2,0.0744217586517334
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,1,1,power_law_1.2,0.09306112289428711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,1,1,power_law_1.2,0.13361472129821778
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,1,1,power_law_1.2,0.13850367546081543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,1,1,power_law_1.2,0.1415443229675293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,1,1,power_law_1.2,0.14485183715820313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,1,1,power_law_1.2,0.14478015899658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,1,1,power_law_1.2,0.14682368278503416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,1,1,power_law_1.2,0.16378303527832033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,1,1,power_law_1.2,0.16683008193969726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,1,1,power_law_1.2,0.17227264404296877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,1,1,power_law_1.2,0.17916799545288087
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,1,1,power_law_1.2,0.18068927764892578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,1,1,power_law_1.2,0.1879007911682129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,1,1,power_law_1.2,0.1950886344909668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,1,1,power_law_1.2,0.21355648040771485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,1,1,power_law_1.2,0.24920448303222656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,1,1,power_law_1.2,0.28237247467041016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,1,1,power_law_1.2,0.35228927612304684
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,1,1,power_law_1.2,2.0981369018554688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,1,1,power_law_1.2,0.42632255554199217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,1,1,power_law_1.2,0.7200345611572265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,1,1,power_law_1.2,0.5717811203002929
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,1,1,power_law_1.2,2.7495941162109374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,1,1,power_law_1.2,1.014211196899414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,1,1,power_law_1.2,0.03491775989532471
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,1,1,power_law_1.2,0.04105279922485351
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,1,1,power_law_1.2,0.046190719604492184
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,1,1,power_law_1.2,0.051459841728210456
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,1,1,power_law_1.2,0.05735167980194091
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,1,1,power_law_1.2,0.07882559776306153
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,1,1,power_law_1.2,0.08258624076843261
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,1,1,power_law_1.2,0.0852678394317627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,1,1,power_law_1.2,0.0871993637084961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,1,1,power_law_1.2,0.08792896270751953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,1,1,power_law_1.2,0.09010368347167969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,1,1,power_law_1.2,0.09360256195068359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,1,1,power_law_1.2,0.0951046371459961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,1,1,power_law_1.2,0.10194944381713866
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,1,1,power_law_1.2,3.391688232421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,1,1,power_law_1.2,0.10694463729858399
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,1,1,power_law_1.2,0.1112665557861328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,1,1,power_law_1.2,1.3067686462402344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,1,1,power_law_1.2,0.12260543823242187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,1,1,power_law_1.2,0.14667072296142578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,1,1,power_law_1.2,0.1737708854675293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,1,1,power_law_1.2,0.2926259231567383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,1,1,power_law_1.2,0.2345350456237793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,1,1,power_law_1.2,0.4222419357299804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,1,1,power_law_1.2,0.5558188629150391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,1,1,power_law_1.2,0.8142176055908203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,1,1,power_law_1.2,1.603173828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,1,1,power_law_1.2,1.0656947326660156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.01,0.07800127983093261
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.01,0.12496191978454589
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.01,0.22543104171752928
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.01,0.3275289535522461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.01,0.4139097595214844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.01,0.631773452758789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.01,0.6749683380126953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.01,0.6860249328613282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.01,0.7209401702880859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.01,0.7249177551269531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.01,0.7822214508056641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.01,0.838307876586914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.01,0.8482701110839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.01,0.8703871917724608
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.01,0.8980665588378907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.01,0.912572784423828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,1,1,power_law_1.2,1.5779980468750001
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.01,0.9508972930908204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.01,1.0454732513427734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.01,1.1453388977050782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.01,1.3826130676269532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.01,1.6551449584960938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.01,2.188662414550781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.01,2.7348654174804685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.01,3.8612069702148437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.01,4.999284362792968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.01,7.264518432617187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,1,1,power_law_1.2,2.1035302734375003
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.01,0.05705344200134277
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.01,9.553961181640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.01,0.08776127815246582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.01,0.18779008865356445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.01,0.24134399414062502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.01,0.35599361419677733
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.01,0.3761427307128906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.01,0.3862432098388672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.01,0.13519743919372557
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.01,0.3975852966308594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.01,0.4101004791259766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,1,1,power_law_1.01,11.830378417968749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.01,0.4462035369873047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.01,0.45966335296630856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.01,0.47056510925292966
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.01,0.47376831054687496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.01,0.4486489486694336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.01,0.491960334777832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.01,0.5186387252807617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.01,0.3974406433105469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.01,0.5501862335205078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.01,0.6251667022705079
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.01,0.7199910736083984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.01,0.8772563171386718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.01,1.0476422119140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.01,1.4014393615722658
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.01,1.7573907470703127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.01,2.475541687011719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.01,3.1926763916015624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,1,1,power_law_1.01,3.9024411010742184
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.01,0.0472211217880249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.01,0.11636287689208984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.01,0.06291520118713378
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.01,0.1434227180480957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.01,0.21548992156982422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.01,0.22987775802612304
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.01,0.08672896385192871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.01,0.23613887786865234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.01,0.24374208450317383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.01,0.24444095611572267
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.01,0.2548051261901855
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.01,0.2587116813659668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.01,0.2621875190734863
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.01,0.2732352066040039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,1,1,power_law_1.2,2.62125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.01,0.2841555213928223
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.01,0.2887590408325195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.01,0.307458553314209
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.01,0.34307838439941407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.01,0.47119617462158203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.01,0.5764275360107421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.01,0.38414592742919923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.01,0.7681510162353515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.01,0.9663699340820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.01,1.3820632934570312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.01,1.8273362731933596
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,1,1,power_law_1.2,0.07993216037750243
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.01,2.700869140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,1,1,power_law_1.2,0.22103424072265626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,1,1,power_law_1.2,0.3032716751098633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,1,1,power_law_1.2,0.38476097106933593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,1,1,power_law_1.2,0.12836992263793945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,1,1,power_law_1.2,0.6449472045898437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,1,1,power_law_1.2,0.6849785614013671
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,1,1,power_law_1.2,0.7091814422607422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,1,1,power_law_1.2,0.7332057952880859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,1,1,power_law_1.2,0.7390476989746093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,1,1,power_law_1.2,0.7892415618896484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,1,1,power_law_1.2,0.8561011505126952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,1,1,power_law_1.2,0.8587744140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,1,1,power_law_1.2,0.8789817810058593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,1,1,power_law_1.2,0.9094866943359374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,1,1,power_law_1.2,0.9270726776123046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,1,1,power_law_1.2,0.9557241821289063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,1,1,power_law_1.2,1.070163803100586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,1,1,power_law_1.2,1.1747161865234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,1,1,power_law_1.2,1.4281228637695311
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,1,1,power_law_1.2,1.6732978820800781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,1,1,power_law_1.2,2.2250341796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,1,1,power_law_1.2,2.780147705078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.01,3.5811059570312502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.01,4.464301452636719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,1,1,power_law_1.2,3.905758056640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,1,1,power_law_1.2,5.033622436523437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,1,1,power_law_1.2,0.055973119735717766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,1,1,power_law_1.2,0.0868620777130127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,1,1,power_law_1.2,0.13166144371032715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,1,1,power_law_1.2,0.1736390495300293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,1,1,power_law_1.2,0.22628416061401366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,1,1,power_law_1.2,0.3598086547851563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,1,1,power_law_1.2,0.3790041732788086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,1,1,power_law_1.2,0.3910879898071289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,1,1,power_law_1.2,0.4016377639770508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,1,1,power_law_1.2,0.40068801879882815
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,1,1,power_law_1.2,0.4075788879394532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,1,1,power_law_1.2,0.4491654586791992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,1,1,power_law_1.2,0.45404735565185544
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,1,1,power_law_1.2,0.4672422409057617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,1,1,power_law_1.2,7.329785766601563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,1,1,power_law_1.2,0.4791596984863281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,1,1,power_law_1.2,0.48299518585205076
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,1,1,power_law_1.2,0.4964985656738281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,1,1,power_law_1.2,0.5255078506469727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,1,1,power_law_1.2,0.5636345672607421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,1,1,power_law_1.2,0.6463609313964843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,1,1,power_law_1.2,0.7269881439208985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,1,1,power_law_1.2,0.8870777893066407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,1,1,power_law_1.2,1.0685305786132813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,1,1,power_law_1.2,1.4166117858886718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,1,1,power_law_1.2,1.7802648925781248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,1,1,power_law_1.2,9.698604125976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,1,1,power_law_1.2,2.4847967529296873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,1,1,power_law_1.2,0.04749120235443115
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,1,1,power_law_1.2,0.06214784145355225
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,1,1,power_law_1.2,0.084017915725708
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,1,1,power_law_1.2,0.1079212760925293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,1,1,power_law_1.2,0.13463808059692384
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,1,1,power_law_1.2,0.21879360198974612
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,1,1,power_law_1.2,0.23299711227416992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,1,1,power_law_1.2,0.24092096328735352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,1,1,power_law_1.2,0.24904767990112306
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,1,1,power_law_1.2,0.24882944107055663
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,1,1,power_law_1.2,0.255743350982666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,1,1,power_law_1.2,0.2660000038146973
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,1,1,power_law_1.2,0.27049856185913085
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,1,1,power_law_1.2,0.2829209518432617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,1,1,power_law_1.2,0.29638336181640623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,1,1,power_law_1.2,0.3046329689025879
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,1,1,power_law_1.2,11.921914062499999
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,1,1,power_law_1.2,3.2016250610351564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,1,1,power_law_1.2,0.33013118743896486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,1,1,power_law_1.2,0.3859379196166992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,1,1,power_law_1.2,0.44605312347412107
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,1,1,power_law_1.2,0.583299217224121
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,1,1,power_law_1.2,0.7200691223144531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,1,1,power_law_1.2,3.922633666992188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,1,1,power_law_1.2,1.004853744506836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,1,1,power_law_1.2,1.3102642822265627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,1,1,power_law_1.2,1.898011474609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,1,1,power_law_1.2,2.4741912841796876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.01,0.14950719833374024
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.01,0.30662975311279295
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.01,0.5085491180419922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.01,0.8169734191894531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.01,1.1207532501220703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.01,1.8087884521484376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.01,1.9229664611816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.01,1.9949562072753906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.01,2.063642883300781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.01,2.110254669189453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.01,2.223852844238281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.01,2.354394836425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.01,2.4096812438964843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.01,2.5266246032714843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.01,2.5668069458007814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,1,1,power_law_1.2,3.63476806640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.01,2.5855474853515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.01,2.746130676269531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.01,3.0840792846679688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.01,3.32166015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.01,3.860680236816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.01,4.5584564208984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.01,5.8217492675781255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.01,7.192901000976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.01,10.053702392578126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.01,12.78931884765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.01,18.395888671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,1,1,power_law_1.2,4.802467956542968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.01,0.11087103843688965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.01,0.16883712768554687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.01,0.28349119186401367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.01,0.4377804946899414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.01,0.5941158294677734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.01,0.9457401275634766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.01,0.9906028747558594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.01,24.2294091796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.01,1.027293472290039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.01,1.0583948516845703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.01,1.072509460449219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.01,1.0986585235595703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.01,1.1241190338134766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.01,1.2014470672607422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.01,1.2549727630615233
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.01,1.2612134552001952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.01,1.2478828430175781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.01,1.3002093505859373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.01,1.3800512695312501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.01,1.4207571411132813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.01,1.540394287109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.01,1.727313232421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.01,2.054373779296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.01,2.376375732421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,1,1,power_law_1.01,30.5172314453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.01,3.1013473510742187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.01,3.873229370117188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.01,5.299615478515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.01,6.767294311523438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,1,1,power_law_1.01,8.308502197265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,1,1,power_law_1.2,6.004545288085938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.01,0.07975808143615723
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.01,0.35155071258544923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.01,0.5680889511108399
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.01,0.11203712463378905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.01,0.6020876693725585
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.01,0.62574462890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.01,0.1700057601928711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.01,0.6474361419677734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.01,0.6565190124511718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.01,0.6745753479003906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.01,0.6903750610351562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.01,0.6949132537841797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.01,0.7270630645751953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.01,0.7361746978759766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.01,0.7376998138427735
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.01,0.2603225517272949
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.01,0.7764793395996094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.01,0.8498655700683594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.01,0.9070265960693359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.01,1.0397401428222657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.01,1.2167110443115234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.01,1.5538809204101562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.01,1.9029592895507812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.01,2.657229309082031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.01,3.3950872802734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,1,1,power_law_1.2,0.15002495765686036
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,1,1,power_law_1.2,0.316499195098877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,1,1,power_law_1.2,0.5034457778930663
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.01,4.942867736816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,1,1,power_law_1.2,0.7486803436279297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,1,1,power_law_1.2,1.7963424682617188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.01,6.637744750976563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,1,1,power_law_1.2,1.8978970336914063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,1,1,power_law_1.2,2.0144422912597655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,1,1,power_law_1.2,1.0296275329589843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,1,1,power_law_1.01,8.232623291015624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,1,1,power_law_1.2,2.0989535522460936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,1,1,power_law_1.2,2.1811622619628905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,1,1,power_law_1.2,2.3052748107910155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,1,1,power_law_1.2,2.5213772583007814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,1,1,power_law_1.2,2.4280787658691407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,1,1,power_law_1.2,2.5604876708984374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,1,1,power_law_1.2,2.6712908935546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,1,1,power_law_1.2,2.721166687011719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,1,1,power_law_1.2,2.851658935546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,1,1,power_law_1.2,3.1116006469726565
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,1,1,power_law_1.2,3.365314636230469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,1,1,power_law_1.2,4.000635375976563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,1,1,power_law_1.2,4.646490783691407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,1,1,power_law_1.2,5.9830078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,1,1,power_law_1.2,7.324631958007812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,1,1,power_law_1.2,10.07620361328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,1,1,power_law_1.2,13.289825439453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,1,1,power_law_1.2,0.11159232139587402
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,1,1,power_law_1.2,0.16781503677368165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,1,1,power_law_1.2,0.273306884765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,1,1,power_law_1.2,0.39452415466308594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,1,1,power_law_1.2,0.5512217712402344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,1,1,power_law_1.2,0.9320031738281249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,1,1,power_law_1.2,0.9807347106933595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,1,1,power_law_1.2,1.0375174713134765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,1,1,power_law_1.2,1.0731097412109374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,1,1,power_law_1.2,1.099425277709961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,1,1,power_law_1.2,1.1296927642822265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,1,1,power_law_1.2,19.846573486328126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,1,1,power_law_1.2,1.1421004486083983
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,1,1,power_law_1.2,1.2507417297363281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,1,1,power_law_1.2,1.2653011322021483
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,1,1,power_law_1.2,1.2932972717285156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,1,1,power_law_1.2,1.3172186279296876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,1,1,power_law_1.2,1.344385223388672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,1,1,power_law_1.2,1.3966009521484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,1,1,power_law_1.2,1.4546560668945312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,1,1,power_law_1.2,1.6039181518554688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,1,1,power_law_1.2,1.7585516357421873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,1,1,power_law_1.2,2.1242393493652343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,1,1,power_law_1.2,2.4407058715820313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,1,1,power_law_1.2,3.1957177734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,1,1,power_law_1.2,3.9111218261718745
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,1,1,power_law_1.2,26.192446289062502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,1,1,power_law_1.2,5.420684814453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,1,1,power_law_1.2,0.07940800189971924
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,1,1,power_law_1.2,0.10801983833312989
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,1,1,power_law_1.2,0.1658790397644043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,1,1,power_law_1.2,0.23433088302612304
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,1,1,power_law_1.2,0.3247385787963867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,1,1,power_law_1.2,0.5597907257080078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,1,1,power_law_1.2,0.5969305419921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,1,1,power_law_1.2,0.6353766250610351
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,1,1,power_law_1.2,0.6560896301269532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,1,1,power_law_1.2,32.382783203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,1,1,power_law_1.2,0.6744032287597657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,1,1,power_law_1.2,6.862905883789063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,1,1,power_law_1.2,0.696460189819336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,1,1,power_law_1.2,0.7069145965576171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,1,1,power_law_1.2,0.7283840179443359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,1,1,power_law_1.2,0.7414937591552734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,1,1,power_law_1.2,8.35057373046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,1,1,power_law_1.2,0.7718534088134765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,1,1,power_law_1.2,0.8334585571289063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,1,1,power_law_1.2,0.7915155029296874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,1,1,power_law_1.2,0.9173407745361327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,1,1,power_law_1.2,1.2351289367675782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,1,1,power_law_1.2,1.0122700500488282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,1,1,power_law_1.2,1.4810470581054687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,1,1,power_law_1.2,2.004864044189453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,1,1,power_law_1.2,2.5535763549804686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,1,1,power_law_1.2,3.7550289916992186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,1,1,power_law_1.2,4.929150695800781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,1,1,power_law_1.01,0.14494784355163576
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,1,1,power_law_1.01,0.28368831634521485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.01,0.49340095520019533
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.01,0.9036601257324218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.01,1.6220889282226563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.01,2.2768595886230467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.01,2.6053811645507814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.01,3.3778335571289064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.01,3.554983825683594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,1,1,power_law_1.2,7.327401123046874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.01,3.716816711425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.01,3.842085266113281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.01,3.896458740234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.01,4.072400512695312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.01,4.285882873535156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.01,4.809744567871094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.01,4.76650634765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.01,5.061703796386719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.01,5.3863549804687505
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.01,5.985049438476563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.01,6.508638305664062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,1,1,power_law_1.2,9.735422973632812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.01,7.6682763671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.01,8.963666381835939
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.01,11.377655029296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.01,13.855137939453124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.01,19.2125732421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.01,0.09486528396606445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.01,0.15254847526550291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.01,0.2682489585876465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.01,0.48166656494140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.01,0.8462322998046876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.01,24.09337158203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.01,1.1967321777343751
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.01,1.3862503051757813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.01,1.756988830566406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.01,1.838880615234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.01,1.917900848388672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.01,1.9781228637695314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.01,2.0045510864257814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.01,2.0685638427734374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,1,1,power_law_1.01,29.3592431640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.01,2.132204132080078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.01,2.221538543701172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.01,2.361802215576172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.01,2.445132141113281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.01,2.532093505859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.01,2.683570556640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.01,2.144780731201172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.01,2.8362835693359374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.01,3.12161865234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.01,3.45666748046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.01,4.131365661621094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.01,4.86262451171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.01,6.160244750976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.01,7.553336181640626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,1,1,power_law_1.01,8.959047241210937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.01,0.10190848350524902
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.01,0.10251392364501952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.01,0.112424955368042
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.01,0.12502911567687988
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.01,0.1382310390472412
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.01,0.10354368209838867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.01,0.14997119903564454
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.01,0.16358463287353514
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.01,0.1757030487060547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.01,0.19948928833007812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.01,0.10570367813110351
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.01,0.2504947280883789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.01,0.27663808822631836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.01,0.32807937622070316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.01,0.39607681274414064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.01,0.4517279815673828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.01,0.5708063888549805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.01,1.0590918731689452
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.01,0.8130169677734376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.01,1.55905029296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.01,2.044613800048828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.01,3.016301574707031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.01,3.9986700439453124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.01,6.008165283203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,1,1,power_law_1.2,0.4932646560668945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,1,1,power_law_1.2,0.9033055877685546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,1,1,power_law_1.2,1.4737350463867187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,1,1,power_law_1.2,2.065404205322266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,1,1,power_law_1.2,2.387170562744141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,1,1,power_law_1.2,3.3025393676757817
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,1,1,power_law_1.2,3.449570617675781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,1,1,power_law_1.2,3.5661413574218748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,1,1,power_law_1.2,3.7050347900390626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,1,1,power_law_1.2,3.893246154785156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,1,1,power_law_1.2,4.298674621582031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,1,1,power_law_1.2,4.598353881835938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,1,1,power_law_1.2,4.0365121459960935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,1,1,power_law_1.2,4.874220886230469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,1,1,power_law_1.2,4.844307250976563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,1,1,power_law_1.2,5.201735229492188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,1,1,power_law_1.2,5.498634643554688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,1,1,power_law_1.2,8.048078002929689
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,1,1,power_law_1.2,9.265867919921876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,1,1,power_law_1.2,6.121359252929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,1,1,power_law_1.2,6.774325561523438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,1,1,power_law_1.2,11.89403564453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,1,1,power_law_1.2,14.317752685546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,1,1,power_law_1.2,19.358758544921876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,1,1,power_law_1.2,0.0936729621887207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,1,1,power_law_1.2,24.6014208984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,1,1,power_law_1.2,0.2669785690307617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,1,1,power_law_1.2,0.14737343788146973
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,1,1,power_law_1.2,0.7696089935302735
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,1,1,power_law_1.2,1.0857510375976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,1,1,power_law_1.2,1.2834860229492189
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,1,1,power_law_1.2,0.47403007507324213
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,1,1,power_law_1.2,1.7196025085449218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,1,1,power_law_1.2,1.7910418701171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,1,1,power_law_1.2,1.846746826171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,1,1,power_law_1.2,1.9143629455566404
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,1,1,power_law_1.2,1.987840576171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,1,1,power_law_1.2,30.242678222656252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,1,1,power_law_1.2,2.0401907348632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,1,1,power_law_1.2,2.101642913818359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,1,1,power_law_1.2,2.1493618774414065
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,1,1,power_law_1.2,2.197729339599609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,1,1,power_law_1.2,2.3748646545410157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,1,1,power_law_1.2,2.5758316040039064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,1,1,power_law_1.2,2.7577133178710938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,1,1,power_law_1.2,2.494526672363281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,1,1,power_law_1.2,2.9472210693359377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,1,1,power_law_1.2,3.2894546508789064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,1,1,power_law_1.2,3.6593414306640626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,1,1,power_law_1.2,4.288296203613281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,1,1,power_law_1.2,4.979368286132813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,1,1,power_law_1.2,6.376298217773437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,1,1,power_law_1.2,7.931675415039062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,1,1,power_law_1.2,0.10322303771972656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,1,1,power_law_1.2,9.237239379882812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,1,1,power_law_1.2,0.10637760162353516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,1,1,power_law_1.2,0.12551551818847656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,1,1,power_law_1.2,0.13807680130004882
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,1,1,power_law_1.2,0.15035584449768066
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,1,1,power_law_1.2,0.1632211112976074
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,1,1,power_law_1.2,0.1756492805480957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,1,1,power_law_1.2,0.20097280502319337
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,1,1,power_law_1.2,0.2505907249450684
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,1,1,power_law_1.2,0.10195072174072266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,1,1,power_law_1.2,0.277270393371582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,1,1,power_law_1.2,0.10239295959472657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,1,1,power_law_1.2,0.39815296173095704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,1,1,power_law_1.2,0.11231807708740234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,1,1,power_law_1.2,0.32816001892089847
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,1,1,power_law_1.2,0.4479673767089844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,1,1,power_law_1.2,0.5735795211791992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,1,1,power_law_1.2,0.8162995147705079
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,1,1,power_law_1.2,1.0557708740234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,1,1,power_law_1.2,1.5466744995117188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,1,1,power_law_1.2,2.0557101440429686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,1,1,power_law_1.2,3.0219628906250002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,1,1,power_law_1.2,3.9994552612304686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,1,1,power_law_1.2,5.962598876953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,power_law_1.01,0.0672659206390381
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.01,0.10120832443237304
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.01,0.16257535934448242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.01,0.3075584030151367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.01,0.47392959594726564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.01,0.49881855010986326
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.01,0.23167808532714843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.01,0.5254380798339844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.01,0.5576947021484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.01,0.5425907135009765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.01,0.5743340682983399
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.01,0.6379084777832031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.01,0.6542681884765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.01,0.6732870483398438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.01,0.6808467102050781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.01,0.6875552368164063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.01,0.7288339233398438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.01,0.7907750701904297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.01,0.8533824157714843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.01,0.9846848297119142
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.01,1.1382157135009767
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.01,1.4483328247070313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.01,1.7932479858398438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.01,2.4767443847656248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.01,3.1583502197265623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.01,4.537081604003906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.01,5.93562255859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,1,1,power_law_1.01,7.315990600585937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,power_law_1.01,0.04105279922485351
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,power_law_1.01,0.11596416473388671
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,power_law_1.01,0.1683475112915039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,power_law_1.01,0.17559808731079102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,power_law_1.01,0.1850604820251465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,power_law_1.01,0.06870975971221924
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,power_law_1.01,0.19145984649658204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,power_law_1.01,0.19468095779418945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,power_law_1.01,0.19876415252685548
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,power_law_1.01,0.05214399814605712
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,power_law_1.01,0.20321023941040037
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,power_law_1.01,0.2077785682678223
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,power_law_1.01,0.21393983840942382
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,power_law_1.01,0.21982143402099608
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,power_law_1.01,0.22497791290283203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,power_law_1.01,0.0917740821838379
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,power_law_1.01,0.23679424285888673
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,power_law_1.01,0.2629753684997559
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.01,0.28948928833007814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.01,0.3440460968017578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.01,0.4063987350463867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.01,0.6684716796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.01,0.5323295974731446
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.01,0.9482169342041015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.01,1.2280818939208984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.01,1.8006156921386718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,1,1,power_law_1.2,0.0845798397064209
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,1,1,power_law_1.2,0.16001792907714846
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,1,1,power_law_1.2,0.2179961585998535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,1,1,power_law_1.2,0.29834367752075197
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.01,2.413690185546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,1,1,power_law_1.2,0.47391742706298834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,1,1,power_law_1.2,0.5044192123413086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,1,1,power_law_1.2,0.5355583953857421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,1,1,power_law_1.2,0.551982078552246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,1,1,power_law_1.2,0.5662208175659179
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,1,1,power_law_1.2,0.5869408035278321
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,1,1,power_law_1.2,0.6612556457519532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,1,1,power_law_1.2,0.673504638671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,1,1,power_law_1.2,0.6881676483154296
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,1,1,power_law_1.2,0.06217984199523926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,1,1,power_law_1.2,0.6974240112304687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,1,1,power_law_1.2,0.7251878356933593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,1,1,power_law_1.2,0.8087238311767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,1,1,power_law_1.2,0.8612780761718749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,1,1,power_law_1.2,1.0127289581298828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,1,1,power_law_1.2,1.1655564880371094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,1,1,power_law_1.2,0.745902099609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,1,1,power_law_1.2,1.4918354797363282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,1,1,power_law_1.2,1.8248876953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,1,1,power_law_1.2,2.5070469665527346
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,1,1,power_law_1.01,3.009415588378906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,1,1,power_law_1.2,3.1926834106445314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,1,1,power_law_1.2,0.04105279922485351
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,1,1,power_law_1.2,0.05261888027191162
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,1,1,power_law_1.2,0.0680787181854248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,1,1,power_law_1.2,0.08607551574707031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,1,1,power_law_1.2,0.10952383995056152
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,1,1,power_law_1.2,4.63084228515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,1,1,power_law_1.2,0.16748287200927733
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,1,1,power_law_1.2,0.17803455352783204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,1,1,power_law_1.2,0.1872902488708496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,1,1,power_law_1.2,0.19303743362426756
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,1,1,power_law_1.2,0.1976416015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,1,1,power_law_1.2,0.2022649574279785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,1,1,power_law_1.2,0.21067712783813475
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,1,1,power_law_1.2,0.21542335510253907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,1,1,power_law_1.2,0.22282432556152343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,1,1,power_law_1.2,6.090785522460938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,1,1,power_law_1.2,7.521184692382812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,1,1,power_law_1.2,0.2309280014038086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,1,1,power_law_1.2,0.23993152618408203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,1,1,power_law_1.2,0.2572403144836426
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,1,1,power_law_1.2,0.29213632583618165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,1,1,power_law_1.2,0.3290316772460938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,1,1,power_law_1.2,0.4204140853881836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,1,1,power_law_1.2,0.5150790405273438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,1,1,power_law_1.2,0.7212384033203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,1,1,power_law_1.2,0.9383020782470703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,1,1,power_law_1.2,1.3788954162597657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,1,1,power_law_1.2,2.8165158081054686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,1,1,power_law_1.2,1.8502642822265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.01,0.2926617622375488
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.01,0.4941958236694336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.01,0.9167520141601562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.01,1.5830323791503906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.01,2.232483215332031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.01,3.684736022949219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.01,3.9337454223632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.01,4.109989624023437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.01,4.234679565429688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,1,1,power_law_1.2,3.7971258544921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.01,4.3167614746093745
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.01,4.530952453613281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.01,5.026023559570312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.01,4.759302978515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.01,5.669285888671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,1,1,power_law_1.2,4.718812255859374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.01,6.260812377929687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.01,6.949151000976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.01,8.240625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.01,6.218474731445313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.01,6.729032592773438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.01,7.825596313476562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.01,9.083002929687499
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.01,11.62606689453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.01,14.313602294921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.01,19.567987060546876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.01,25.266806640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.01,0.17064512252807618
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.01,0.2786393547058105
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.01,0.4915340805053711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.01,37.715036621093745
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.01,0.8310598754882813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.01,1.1857881927490235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.01,1.9354643249511718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.01,2.060072326660156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.01,2.1501228332519533
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.01,2.2185209655761717
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.01,2.249070739746094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.01,2.3038758850097656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.01,2.3655917358398435
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.01,52.4521923828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.01,2.394559326171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.01,2.479917449951172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.01,2.5836978149414063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.01,2.6648065185546876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.01,2.5013529968261716
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.01,2.839399719238281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.01,2.9800326538085935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.01,3.6371142578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.01,3.2971725463867188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.01,4.351506652832031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.01,5.0436376953124995
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.01,6.625038452148438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.01,8.110122680664062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.01,11.283582763671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.01,14.4818115234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,power_law_1.01,0.1558944034576416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,power_law_1.01,0.16015296936035156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,power_law_1.01,0.16658687591552734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,power_law_1.01,0.15625920295715331
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,power_law_1.01,0.18130624771118165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,power_law_1.01,0.19483776092529298
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,power_law_1.01,0.20753152847290038
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,power_law_1.01,0.2215878486633301
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,power_law_1.01,0.23389375686645505
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,power_law_1.01,0.2609657669067383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,power_law_1.01,0.15578623771667482
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,power_law_1.01,0.3134086418151855
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,power_law_1.01,0.3907948684692383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,power_law_1.01,0.3406079864501953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,power_law_1.01,0.47551551818847654
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,power_law_1.01,0.5290950393676758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,power_law_1.01,0.6766854095458984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,power_law_1.01,0.95316162109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.01,1.2395225524902345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.01,1.807316436767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.01,2.3853414916992186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.01,3.517017517089844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.01,4.64568115234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.01,6.952011718750001
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.01,9.293806762695313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,1,1,power_law_1.2,0.4941497421264648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,1,1,power_law_1.2,0.9182310485839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,1,1,power_law_1.2,0.2957478332519531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,1,1,power_law_1.2,1.4484941101074218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,1,1,power_law_1.2,3.613438110351562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,1,1,power_law_1.2,1.9926304626464844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,1,1,power_law_1.2,3.9299188232421876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,1,1,power_law_1.2,4.057976379394531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,1,1,power_law_1.2,4.235730590820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,1,1,power_law_1.2,4.415960998535157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,1,1,power_law_1.2,4.650575866699219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,1,1,power_law_1.2,4.9602899169921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,1,1,power_law_1.2,5.260819702148437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,1,1,power_law_1.2,5.928946533203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,1,1,power_law_1.2,6.60799560546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,1,1,power_law_1.2,7.232490234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,1,1,power_law_1.2,8.538475952148437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,1,1,power_law_1.2,6.471700439453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,1,1,power_law_1.2,8.107132568359376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,1,1,power_law_1.2,9.318547363281251
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,1,1,power_law_1.2,6.946461791992188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,1,1,power_law_1.2,11.9914990234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,1,1,power_law_1.2,14.690240478515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,1,1,power_law_1.2,20.609306640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,1,1,power_law_1.2,26.909189453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,1,1,power_law_1.2,0.17071807861328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,1,1,power_law_1.2,0.2757107162475586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,1,1,power_law_1.2,0.48751102447509764
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,1,1,power_law_1.2,0.7593106842041015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,1,1,power_law_1.2,1.0691519927978514
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,1,1,power_law_1.2,1.8996780395507813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,1,1,power_law_1.2,2.0599366760253908
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,1,1,power_law_1.2,42.324716796874995
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,1,1,power_law_1.2,2.1256781005859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,1,1,power_law_1.2,2.2776934814453123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,1,1,power_law_1.2,2.208643798828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,1,1,power_law_1.2,2.3450271606445314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,1,1,power_law_1.2,2.4053868103027343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,1,1,power_law_1.2,2.453813781738281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,1,1,power_law_1.2,2.5632000732421876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,1,1,power_law_1.2,2.6115634155273435
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,1,1,power_law_1.2,2.6645547485351564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,1,1,power_law_1.2,2.7747052001953127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,1,1,power_law_1.2,2.962757263183594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,1,1,power_law_1.2,3.1031896972656248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,1,1,power_law_1.2,3.4120364379882817
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,1,1,power_law_1.2,3.7755679321289066
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,1,1,power_law_1.2,4.438897399902343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,1,1,power_law_1.2,5.203645629882812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,1,1,power_law_1.2,56.2016455078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,1,1,power_law_1.2,6.701343383789063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,1,1,power_law_1.2,8.397001342773438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,1,1,power_law_1.2,11.35385498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,1,1,power_law_1.2,14.446414794921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,1,1,power_law_1.2,0.1562713623046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,1,1,power_law_1.2,0.15757760047912597
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,1,1,power_law_1.2,0.1548256015777588
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,1,1,power_law_1.2,0.1670252799987793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,1,1,power_law_1.2,0.18171775817871094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,1,1,power_law_1.2,0.1945292854309082
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,1,1,power_law_1.2,0.20788543701171874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,1,1,power_law_1.2,0.22039039611816405
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,1,1,power_law_1.2,0.23463232040405274
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,1,1,power_law_1.2,0.26024192810058594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,1,1,power_law_1.2,0.33885822296142576
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,1,1,power_law_1.2,0.39140350341796876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,1,1,power_law_1.2,0.4759494400024414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,1,1,power_law_1.2,0.5296499252319335
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,1,1,power_law_1.2,0.15904319763183594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,1,1,power_law_1.2,0.31290048599243164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,1,1,power_law_1.2,0.6735929870605469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,1,1,power_law_1.2,0.9536544036865233
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,1,1,power_law_1.2,1.8043040466308593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,1,1,power_law_1.2,1.2385363006591796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,1,1,power_law_1.2,2.370738525390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,1,1,power_law_1.2,3.5203314208984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,1,1,power_law_1.2,4.6514584350585935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,1,1,power_law_1.2,6.940380249023437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.01,0.13991744041442872
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.01,0.15308799743652343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.01,0.18895423889160157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.01,0.25025344848632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.01,0.3120966339111328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.01,0.3141318321228027
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.01,0.31729663848876954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.01,0.31573375701904294
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.01,0.3225337600708008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.01,0.33823806762695313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.01,0.35863361358642576
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.01,0.35725376129150394
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,1,1,power_law_1.2,9.2622265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.01,0.36894271850585936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.01,0.4292044830322266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.01,0.4981811141967773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.01,0.3933772659301758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.01,0.5722035217285156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.01,0.7145407867431641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.01,0.863653793334961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.01,1.1821030426025392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.01,1.4967634582519531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.01,2.1464767456054688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.01,2.7928549194335934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.01,4.104057006835937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.01,5.4542346191406255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.01,9.12924072265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.01,0.06959167957305909
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.01,0.09631872177124023
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.01,0.10731904029846191
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.01,0.14626367568969728
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.01,0.15401472091674803
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.01,0.15561216354370117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.01,0.1566310405731201
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.01,0.15560704231262207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.01,0.1567961597442627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.01,0.16099199295043945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.01,0.1604876708984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.01,0.168666877746582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.01,0.1881113624572754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.01,0.19720895767211916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.01,0.2045100784301758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.01,0.21508224487304686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.01,0.25032768249511717
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.01,0.2988287925720215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.01,0.15764991760253905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.01,0.38110271453857425
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.01,13.447952880859376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.01,0.4553433609008789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.01,0.6232556915283203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.01,0.8082406616210938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.01,1.1377734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.01,1.4722630310058595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.01,2.1706533813476563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,2,1,power_law_1.01,17.468641357421877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.01,2.8485760498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,2,1,power_law_1.01,3.5366458129882816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.01,0.03666687965393066
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.01,0.08499391555786133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.01,0.08498944282531738
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.01,0.08502719879150392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.01,0.08458304405212402
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.01,0.08475199699401856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.01,0.0861747169494629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.01,0.08655743598937989
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.01,0.048592638969421384
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.01,0.06163584232330323
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.01,0.08398591995239257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.01,0.08707967758178711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.01,0.08810815811157227
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.01,0.0976255989074707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.01,0.09993472099304199
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.01,0.10438719749450684
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.01,0.11471487998962401
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.01,0.1410316753387451
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.01,0.16679424285888672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.01,0.21684032440185547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.01,0.2691609573364258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.01,0.37301631927490236
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.01,0.46745536804199217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.01,0.6738642883300782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.01,0.8901516723632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,2,1,power_law_1.2,0.11342080116271973
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,2,1,power_law_1.2,0.15072319984436036
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,2,1,power_law_1.2,0.1652556800842285
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,2,1,power_law_1.2,0.25124671936035153
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,2,1,power_law_1.2,0.306375675201416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,2,1,power_law_1.2,0.313057918548584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,2,1,power_law_1.2,0.3155276870727539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,2,1,power_law_1.2,0.31345216751098637
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,2,1,power_law_1.2,0.3177356719970703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,2,1,power_law_1.2,0.32886207580566407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,2,1,power_law_1.2,0.3602163314819336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,2,1,power_law_1.2,0.36358463287353515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,2,1,power_law_1.2,0.3652076721191406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,2,1,power_law_1.2,0.39121150970458984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,2,1,power_law_1.2,0.4211763381958008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,2,1,power_law_1.2,0.5148646545410156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,2,1,power_law_1.2,0.5751622390747071
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.01,1.3115922546386718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,2,1,power_law_1.2,0.7177375793457031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,2,1,power_law_1.2,0.8705133056640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,2,1,power_law_1.2,1.185445785522461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,2,1,power_law_1.2,1.5005888366699218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,2,1,power_law_1.2,2.1486143493652343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,2,1,power_law_1.2,2.78890380859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.01,1.715692138671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,2,1,power_law_1.2,4.102553100585938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.01,2.146401214599609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,2,1,power_law_1.2,5.546465454101563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,2,1,power_law_1.2,0.071113600730896
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,2,1,power_law_1.2,0.09379776000976563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,2,1,power_law_1.2,0.10076543807983399
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,2,1,power_law_1.2,0.14691200256347656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,2,1,power_law_1.2,0.15153856277465821
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,2,1,power_law_1.2,0.15653056144714356
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,2,1,power_law_1.2,0.15928640365600585
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,2,1,power_law_1.2,0.1566220760345459
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,2,1,power_law_1.2,0.1611891174316406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,2,1,power_law_1.2,0.16107391357421874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,2,1,power_law_1.2,0.16429567337036133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,2,1,power_law_1.2,0.1643244743347168
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,2,1,power_law_1.2,0.16979072570800782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,2,1,power_law_1.2,0.18939008712768554
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,2,1,power_law_1.2,0.20386943817138672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,2,1,power_law_1.2,0.21211584091186522
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,2,1,power_law_1.2,0.21448192596435547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,2,1,power_law_1.2,0.24758783340454102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,2,1,power_law_1.2,0.29931776046752934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,2,1,power_law_1.2,0.36997505187988283
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,2,1,power_law_1.2,0.45528320312500004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,2,1,power_law_1.2,9.312132568359376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,2,1,power_law_1.2,0.6284870529174805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,2,1,power_law_1.2,0.8021984100341797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,2,1,power_law_1.2,1.1396864318847657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,2,1,power_law_1.2,13.755758056640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,2,1,power_law_1.2,1.4718048095703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,2,1,power_law_1.2,2.1791590881347656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,2,1,power_law_1.2,0.03686079978942871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,2,1,power_law_1.2,0.048604798316955564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,2,1,power_law_1.2,0.05343552112579346
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,2,1,power_law_1.2,0.08375231742858887
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,2,1,power_law_1.2,0.08303104400634766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,2,1,power_law_1.2,0.08523200035095216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,2,1,power_law_1.2,0.08525055885314942
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,2,1,power_law_1.2,0.08435839653015137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,2,1,power_law_1.2,17.555780029296876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,2,1,power_law_1.2,0.08544896125793458
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,2,1,power_law_1.2,0.08603327751159669
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,2,1,power_law_1.2,0.08646656036376953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,2,1,power_law_1.2,0.08780672073364257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,2,1,power_law_1.2,0.08888575553894043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,2,1,power_law_1.2,0.09805376052856445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,2,1,power_law_1.2,0.10836095809936523
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,2,1,power_law_1.2,2.9108056640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,2,1,power_law_1.2,0.10382911682128906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,2,1,power_law_1.2,0.11749055862426758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,2,1,power_law_1.2,0.21864255905151367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,2,1,power_law_1.2,3.518671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,2,1,power_law_1.2,0.144520959854126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,2,1,power_law_1.2,0.1699135971069336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,2,1,power_law_1.2,0.2695071983337402
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,2,1,power_law_1.2,0.37748607635498044
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,2,1,power_law_1.2,0.4830003356933593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,2,1,power_law_1.2,0.7038579559326171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,2,1,power_law_1.2,0.9022176361083984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.01,0.15122176170349122
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.01,0.24755199432373046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.01,0.30265279769897463
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.01,0.42685886383056637
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.01,0.5064633560180665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.01,0.5132051086425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.01,0.5281887817382812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.01,0.5214342498779297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.01,0.5273561477661133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.01,0.5381753540039063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.01,0.563493766784668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.01,0.5977612686157227
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.01,0.6426092529296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.01,0.6944595336914062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.01,0.725599365234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.01,0.7666912078857422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.01,0.9174304199218749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.01,1.1699142456054688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.01,1.428613739013672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.01,1.9713818359375002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.01,2.514421691894531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,2,1,power_law_1.2,1.3226002502441407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.01,3.622272644042969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.01,4.725780334472656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.01,6.972581176757812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.01,9.402915039062501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,2,1,power_law_1.2,1.7879493713378907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.01,15.796422119140624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.01,0.08919103622436524
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.01,0.1341926383972168
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.01,0.16269760131835936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.01,0.23777727127075193
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.01,0.2544550323486328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.01,0.2564031982421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.01,0.2576902389526367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.01,0.25451711654663084
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.01,0.2551807975769043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.01,0.26183935165405275
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.01,0.26425344467163087
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.01,0.2611903953552246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.01,0.2676780891418457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.01,0.2770924758911133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.01,0.28165632247924804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.01,0.2898515129089355
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.01,0.31735424041748045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.01,0.3775385665893555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.01,0.43917697906494146
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,2,1,power_law_1.2,2.2027801513671874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.01,0.5590374374389648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.01,0.680769271850586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.01,0.9378406524658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.01,1.2157190704345704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.01,23.97864990234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.01,1.7346853637695312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.01,2.270708465576172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.01,3.321089782714844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,2,1,power_law_1.01,32.3789599609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.01,4.424043579101562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,2,1,power_law_1.01,5.468873901367187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.01,0.04740799903869629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.01,0.13202688217163086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.01,0.13217344284057617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.01,0.1324492835998535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.01,0.1299724769592285
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.01,0.13081151962280274
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.01,0.13347583770751953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.01,0.1341913604736328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.01,0.0761849594116211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.01,0.0893273639678955
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.01,0.1288377571105957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.01,0.1335315227508545
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.01,0.13819775581359864
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.01,0.14980159759521486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.01,0.1520582389831543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.01,0.15919424057006837
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.01,0.17737152099609374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.01,0.21951040267944336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.01,0.2572166442871094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.01,0.33622207641601565
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.01,0.41965057373046877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.01,0.5755487823486328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.01,0.7417030334472656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.01,1.0538253021240234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.01,1.407394561767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,2,1,power_law_1.2,0.15158207893371584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,2,1,power_law_1.2,0.23338367462158205
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,2,1,power_law_1.2,0.27071615219116213
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,2,1,power_law_1.2,0.4249331283569336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,2,1,power_law_1.2,0.48234432220458984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,2,1,power_law_1.2,0.5245081710815429
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,2,1,power_law_1.2,0.5341036987304688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,2,1,power_law_1.2,0.5264441680908203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,2,1,power_law_1.2,0.5413932800292969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,2,1,power_law_1.2,0.5424217605590821
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,2,1,power_law_1.2,0.5758457565307618
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,2,1,power_law_1.2,0.6107136154174805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,2,1,power_law_1.2,0.6318751907348632
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,2,1,power_law_1.2,0.6918035125732421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,2,1,power_law_1.2,0.727136001586914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,2,1,power_law_1.2,0.7810546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,2,1,power_law_1.2,0.9100717163085937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.01,2.0407206726074216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,2,1,power_law_1.2,1.1749900817871093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,2,1,power_law_1.2,1.4390042114257813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,2,1,power_law_1.2,1.9699891662597657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,2,1,power_law_1.2,2.5232826232910157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,2,1,power_law_1.2,3.628749389648438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.01,2.6968045043945312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,2,1,power_law_1.2,4.7176339721679685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,2,1,power_law_1.2,7.027749633789062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.01,3.3364007568359377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,2,1,power_law_1.2,9.677713623046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,2,1,power_law_1.2,0.08820096015930176
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,2,1,power_law_1.2,0.12934080123901368
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,2,1,power_law_1.2,0.14290111541748046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,2,1,power_law_1.2,0.23719360351562502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,2,1,power_law_1.2,0.24443775177001953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,2,1,power_law_1.2,0.25632511138916014
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,2,1,power_law_1.2,0.25773504257202146
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,2,1,power_law_1.2,0.2543424034118652
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,2,1,power_law_1.2,0.26001663208007814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,2,1,power_law_1.2,0.2628409576416016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,2,1,power_law_1.2,0.26472959518432615
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,2,1,power_law_1.2,0.26729856491088866
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,2,1,power_law_1.2,0.2641196823120117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,2,1,power_law_1.2,0.2781171226501465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,2,1,power_law_1.2,0.28840959548950196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,2,1,power_law_1.2,0.29897151947021483
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,2,1,power_law_1.2,0.32932544708251954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,2,1,power_law_1.2,0.37528190612792967
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,2,1,power_law_1.2,0.44038848876953124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,2,1,power_law_1.2,0.5513945770263672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,2,1,power_law_1.2,0.6823219299316406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,2,1,power_law_1.2,16.290438232421874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,2,1,power_law_1.2,0.9446790313720703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,2,1,power_law_1.2,1.1839091491699218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,2,1,power_law_1.2,1.7266227722167968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,2,1,power_law_1.2,2.2546099853515624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,2,1,power_law_1.2,24.452924804687502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,2,1,power_law_1.2,3.3424972534179687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,2,1,power_law_1.2,0.047157759666442874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,2,1,power_law_1.2,0.07230144023895264
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,2,1,power_law_1.2,0.08212160110473633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,2,1,power_law_1.2,0.1293977642059326
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,2,1,power_law_1.2,0.12720512390136718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,2,1,power_law_1.2,0.13256959915161132
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,2,1,power_law_1.2,0.13257472038269041
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,2,1,power_law_1.2,0.13034560203552245
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,2,1,power_law_1.2,0.13314623832702638
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,2,1,power_law_1.2,0.13368831634521486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,2,1,power_law_1.2,0.1347475242614746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,2,1,power_law_1.2,33.15047607421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,2,1,power_law_1.2,0.13808832168579102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,2,1,power_law_1.2,0.14990912437438966
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,2,1,power_law_1.2,0.15740544319152833
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,2,1,power_law_1.2,0.16576383590698243
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,2,1,power_law_1.2,4.434294433593751
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,2,1,power_law_1.2,0.2223526382446289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,2,1,power_law_1.2,0.1357376003265381
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,2,1,power_law_1.2,0.18162176132202149
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,2,1,power_law_1.2,5.574615478515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,2,1,power_law_1.2,0.25962623596191403
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,2,1,power_law_1.2,0.3387251281738281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,2,1,power_law_1.2,0.41890625000000004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,2,1,power_law_1.2,0.595500144958496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,2,1,power_law_1.2,0.7444115447998046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,2,1,power_law_1.2,1.0842527770996093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,2,1,power_law_1.2,1.4050668334960936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.01,0.09830656051635742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.01,0.14073280334472654
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.01,0.2871686363220215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.01,0.5092416000366211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.01,0.730571517944336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.01,0.9327507019042969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.01,1.271717758178711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.01,1.332635498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.01,1.3636773681640624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.01,1.379617919921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.01,1.4516038513183593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.01,1.5283506774902345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.01,1.5668812561035157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.01,1.6706233215332031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.01,1.7472396850585938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.01,1.7621069335937498
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.01,1.8442469787597655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,2,1,power_law_1.2,2.0951052856445314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.01,1.964409637451172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.01,2.11510009765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.01,2.3980307006835937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.01,2.675722961425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.01,3.2443475341796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.01,3.883930969238281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.01,5.170640869140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.01,6.507355346679688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.01,9.147000732421876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.01,11.85919189453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.01,0.06720064163208009
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.01,0.09392831802368165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.01,0.15569791793823243
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.01,0.2755014419555664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.01,0.4027020645141602
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.01,0.525976333618164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.01,0.7085209655761718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,2,1,power_law_1.01,14.530989990234374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.01,0.7467488098144531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.01,0.7593305969238282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.01,0.7705951690673828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.01,0.8035686492919922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.01,0.8284210968017579
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.01,0.8389472198486327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.01,0.8537561798095703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.01,0.9320620727539062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,2,1,power_law_1.2,2.7420318603515623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.01,0.9378765106201172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.01,0.9671981048583984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.01,1.0057727813720703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.01,1.0606848144531251
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.01,1.1505203247070312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.01,1.2424301147460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.01,1.432439727783203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.01,1.6448390197753906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.01,2.0744006347656248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.01,2.533524475097656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.01,3.376712341308594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.01,4.2501550292968755
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,2,1,power_law_1.01,5.1398828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,2,1,power_law_1.2,3.396184387207031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.01,0.0705734395980835
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.01,0.08869183540344239
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.01,0.2470368003845215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.01,0.31160064697265627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.01,0.4326284790039063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.01,0.45640190124511715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.01,0.4687673568725586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.01,0.11752832412719727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.01,0.47406654357910155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.01,0.495715217590332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.01,0.5133087921142578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.01,0.5223046493530273
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.01,0.5325395202636718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.01,0.5560332870483398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.01,0.5924300765991211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.01,0.5611487960815429
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.01,0.17391296386718752
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.01,0.6405817413330078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.01,0.6971225738525391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.01,0.8011827087402343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.01,0.9075161743164063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.01,1.1425881958007813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.01,1.3873440551757814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.01,1.9270637512207032
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.01,2.422848663330078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,2,1,power_law_1.2,0.09726911544799804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,2,1,power_law_1.2,0.13852864265441894
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,2,1,power_law_1.2,0.282040958404541
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,2,1,power_law_1.2,0.44070655822753907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,2,1,power_law_1.2,0.6504537963867187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,2,1,power_law_1.2,0.8641600036621094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,2,1,power_law_1.2,1.2479341125488281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,2,1,power_law_1.2,1.3016038513183594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,2,1,power_law_1.2,1.3314189147949218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,2,1,power_law_1.2,1.3909797668457031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,2,1,power_law_1.2,1.446688690185547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,2,1,power_law_1.2,1.5181709289550782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,2,1,power_law_1.2,1.5893740844726563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,2,1,power_law_1.2,1.7315930175781251
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,2,1,power_law_1.2,1.7687794494628908
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,2,1,power_law_1.01,3.508040466308594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,2,1,power_law_1.01,4.581633911132813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,2,1,power_law_1.2,1.8146028137207033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,2,1,power_law_1.2,1.8894744873046876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,2,1,power_law_1.01,5.8406481933593755
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,2,1,power_law_1.2,2.043390655517578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,2,1,power_law_1.2,2.1803033447265627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,2,1,power_law_1.2,2.4882528686523435
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,2,1,power_law_1.2,2.743857421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,2,1,power_law_1.2,3.3564697265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,2,1,power_law_1.2,4.0089419555664065
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,2,1,power_law_1.2,5.283495483398438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,2,1,power_law_1.2,6.624752197265624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,2,1,power_law_1.2,0.06778687953948974
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,2,1,power_law_1.2,9.292781982421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,2,1,power_law_1.2,0.09377856254577636
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,2,1,power_law_1.2,0.15604031562805176
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,2,1,power_law_1.2,0.25588863372802734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,2,1,power_law_1.2,0.3723052978515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,2,1,power_law_1.2,0.5032479858398438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,2,1,power_law_1.2,0.695884780883789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,2,1,power_law_1.2,11.989682617187501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,2,1,power_law_1.2,0.7269363403320312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,2,1,power_law_1.2,0.7461862182617187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,2,1,power_law_1.2,0.7740422058105468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,2,1,power_law_1.2,0.7949664306640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,2,1,power_law_1.2,0.8183884429931642
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,2,1,power_law_1.2,0.838897247314453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,2,1,power_law_1.2,0.8681644439697266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,2,1,power_law_1.2,0.9405401611328126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,2,1,power_law_1.2,0.964111328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,2,1,power_law_1.2,0.9857465362548827
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,2,1,power_law_1.2,1.0494226837158203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,2,1,power_law_1.2,1.096783981323242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,2,1,power_law_1.2,14.682061767578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,2,1,power_law_1.2,1.2020838165283203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,2,1,power_law_1.2,1.288543395996094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,2,1,power_law_1.2,1.4973139953613281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,2,1,power_law_1.2,1.7255186462402343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,2,1,power_law_1.2,2.1224044799804687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,2,1,power_law_1.2,2.5846163940429685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,2,1,power_law_1.2,0.07039231777191161
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,2,1,power_law_1.2,3.4473165893554687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,2,1,power_law_1.2,0.08799615859985352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,2,1,power_law_1.2,4.3694189453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,2,1,power_law_1.2,0.11419712066650392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,2,1,power_law_1.2,0.16331968307495118
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,2,1,power_law_1.2,0.22745983123779298
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,2,1,power_law_1.2,5.17579345703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,2,1,power_law_1.2,0.2941574478149414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,2,1,power_law_1.2,0.42483070373535153
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,2,1,power_law_1.2,0.4459686279296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,2,1,power_law_1.2,0.4601267242431641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,2,1,power_law_1.2,0.4768403244018555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,2,1,power_law_1.2,0.4956000137329101
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,2,1,power_law_1.2,0.5118368148803711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,2,1,power_law_1.2,0.5277094268798829
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,2,1,power_law_1.2,0.5519961547851563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,2,1,power_law_1.2,0.572745590209961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,2,1,power_law_1.2,0.5915468978881836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,2,1,power_law_1.2,0.6262169647216796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,2,1,power_law_1.2,0.7068345642089844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,2,1,power_law_1.2,0.7802349090576172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,2,1,power_law_1.2,0.9439513397216797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,2,1,power_law_1.2,1.1337983703613281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,2,1,power_law_1.2,1.508062744140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,2,1,power_law_1.2,1.9282380676269533
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,2,1,power_law_1.2,2.733934631347656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.01,0.028028800487518313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.01,0.05375552177429199
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.01,0.0683622407913208
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.01,0.08008319854736327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.01,0.1180339241027832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.01,0.12105024337768555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.01,0.12374784469604491
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.01,0.12668607711791993
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.01,0.12708864212036133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.01,0.1322976016998291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.01,0.1451136016845703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.01,0.14691519737243652
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.01,0.1501356792449951
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.01,0.15354816436767577
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.01,0.15575103759765624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.01,0.16197696685791016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.01,0.19016767501831056
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.01,0.2053772735595703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.01,0.24649536132812502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.01,0.2889913558959961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.01,0.38113662719726565
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.01,0.47347393035888674
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.01,0.6633785247802735
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.01,0.8518508911132813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.01,1.2293049621582033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.01,1.6062471008300783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,2,1,power_law_1.01,1.9842008972167968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.01,0.030933759212493896
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.01,0.03501247882843018
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.01,0.045145602226257325
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.01,0.05758912086486816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.01,0.07246528148651124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.01,0.09350848197937012
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.01,0.09666175842285156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.01,0.09756416320800781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.01,0.09999039649963379
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.01,0.1009708786010742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,2,1,power_law_1.2,3.630474853515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.01,0.10238847732543946
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.01,0.11633343696594238
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.01,0.11723072052001952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.01,0.1191372776031494
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.01,0.12201087951660157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.01,0.12340096473693847
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.01,0.12946751594543457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.01,0.13553088188171386
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.01,0.15052672386169436
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.01,0.1704159927368164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.01,0.20243711471557618
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.01,0.2504832077026367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.01,0.30094720840454103
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.01,0.4088390350341797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.01,0.5168761444091797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.01,0.7296768188476562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.01,0.943249282836914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,2,1,power_law_1.01,1.1501152038574218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.01,0.03536191940307617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.01,0.042780799865722655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.01,0.03778048038482666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.01,0.046197757720947266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.01,0.050469760894775395
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.01,0.06267776012420653
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.01,0.06484864234924317
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.01,0.06579520225524901
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.01,0.06700160026550292
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.01,0.06747968196868896
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.01,0.06964799880981445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.01,0.07109568119049073
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.01,0.07242496013641357
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.01,0.07593984127044677
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.01,0.07921408176422119
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.01,0.08185024261474609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.01,0.08664511680603028
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.01,0.09890048027038574
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.01,0.11439231872558593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.01,0.14421119689941406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.01,0.17334976196289062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.01,0.23608959197998045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.01,0.3037843132019043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.01,0.44620800018310547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.01,0.5879481506347657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.01,0.8845702362060546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.01,1.1937471771240236
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,2,1,power_law_1.01,1.505068817138672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,2,1,power_law_1.2,0.06503935813903808
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,2,1,power_law_1.2,0.07631167888641358
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,2,1,power_law_1.2,0.05239935874938965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,2,1,power_law_1.2,0.11951807975769042
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,2,1,power_law_1.2,0.12175680160522462
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,2,1,power_law_1.2,0.12592448234558104
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,2,1,power_law_1.2,0.12757823944091798
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,2,1,power_law_1.2,0.12788415908813477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,2,1,power_law_1.2,0.1331763172149658
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,2,1,power_law_1.2,0.14643775939941406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,2,1,power_law_1.2,0.14806079864501953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,2,1,power_law_1.2,0.027424640655517578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,2,1,power_law_1.2,0.15146431922912598
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,2,1,power_law_1.2,0.1543616008758545
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,2,1,power_law_1.2,0.03880575895309448
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,2,1,power_law_1.2,0.16379520416259766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,2,1,power_law_1.2,0.15599616050720216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,2,1,power_law_1.2,0.19221824645996094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,2,1,power_law_1.2,0.20613824844360354
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,2,1,power_law_1.2,0.2521331214904785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,2,1,power_law_1.2,0.2903078460693359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,2,1,power_law_1.2,0.38256832122802736
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,2,1,power_law_1.2,0.47741825103759766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,2,1,power_law_1.2,0.6672064208984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,2,1,power_law_1.2,0.8551487731933595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,2,1,power_law_1.2,0.03048320055007935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,2,1,power_law_1.2,0.035762560367584226
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,2,1,power_law_1.2,0.04358399868011474
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,2,1,power_law_1.2,0.0524128007888794
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,2,1,power_law_1.2,0.06942080020904541
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,2,1,power_law_1.2,0.09365951538085937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,2,1,power_law_1.2,0.0962495994567871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,2,1,power_law_1.2,0.09853631973266601
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,2,1,power_law_1.2,0.09994688034057617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,2,1,power_law_1.2,0.10061311721801758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,2,1,power_law_1.2,0.10236800193786622
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,2,1,power_law_1.2,0.11574912071228027
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,2,1,power_law_1.2,0.11698431968688965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,2,1,power_law_1.2,0.1200108814239502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,2,1,power_law_1.2,0.12292799949645997
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,2,1,power_law_1.2,0.12416576385498046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,2,1,power_law_1.2,0.12950336456298828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,2,1,power_law_1.2,0.13721343994140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,2,1,power_law_1.2,0.15309503555297851
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,2,1,power_law_1.2,0.17630592346191407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,2,1,power_law_1.2,0.20273344039916993
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,2,1,power_law_1.2,0.25124095916748046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,2,1,power_law_1.2,0.30595136642456056
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,2,1,power_law_1.2,1.2334886169433594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,2,1,power_law_1.2,0.410753288269043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,2,1,power_law_1.2,0.5152339172363282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,2,1,power_law_1.2,1.6099942016601563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,2,1,power_law_1.2,0.7307405090332031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,2,1,power_law_1.2,0.03572864055633545
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,2,1,power_law_1.2,0.03857408046722412
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,2,1,power_law_1.2,0.04239359855651855
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,2,1,power_law_1.2,0.04488448143005371
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,2,1,power_law_1.2,0.049132161140441895
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,2,1,power_law_1.2,0.06285632133483887
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,2,1,power_law_1.2,0.06506559848785401
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,2,1,power_law_1.2,0.0660038423538208
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,2,1,power_law_1.2,0.06749184131622314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,2,1,power_law_1.2,0.06848000049591065
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,2,1,power_law_1.2,0.07027455806732177
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,2,1,power_law_1.2,0.0727180814743042
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,2,1,power_law_1.2,0.07400767803192139
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,2,1,power_law_1.2,0.07881216049194337
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,2,1,power_law_1.2,0.08269375801086426
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,2,1,power_law_1.2,1.9895610046386718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,2,1,power_law_1.2,0.08652416229248047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,2,1,power_law_1.2,0.9413043212890624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,2,1,power_law_1.2,0.09620160102844239
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,2,1,power_law_1.2,0.11552319526672364
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,2,1,power_law_1.2,0.13743359565734864
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,2,1,power_law_1.2,0.18496639251708985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,2,1,power_law_1.2,0.2327564811706543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,2,1,power_law_1.2,0.33488449096679684
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,2,1,power_law_1.2,1.1534329223632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,2,1,power_law_1.2,0.4421612930297852
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,2,1,power_law_1.2,0.6457766723632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,2,1,power_law_1.2,0.8544601440429688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.01,0.06087296009063721
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.01,0.08104767799377441
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.01,0.12297663688659669
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.01,0.192161922454834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.01,0.2291360092163086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.01,0.341187858581543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.01,0.35739070892333985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.01,0.364582405090332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.01,0.3869094467163086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.01,0.39314624786376956
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.01,0.4155366516113281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.01,0.43775680541992185
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.01,0.44063552856445315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.01,0.46312320709228516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.01,0.4951347351074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.01,0.4975603103637695
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.01,0.5135334396362305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.01,0.5691161727905273
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.01,0.6134815979003906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.01,0.7477894592285156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.01,0.8925145721435547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.01,1.1806073760986329
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.01,1.4756204223632812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.01,2.0814054870605467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,2,1,power_law_1.2,1.26780029296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.01,2.688326416015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.01,3.9021881103515623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.01,5.121154174804688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.01,0.037761919498443604
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.01,0.06053311824798584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.01,0.09326975822448731
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.01,0.12118783950805664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.01,0.1541945552825928
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.01,0.21902080535888674
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.01,0.2265875244140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.01,0.22998720169067383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,2,1,power_law_1.01,6.340183715820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.01,0.23752319335937502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.01,0.23615039825439452
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.01,0.24457279205322266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.01,0.27063488006591796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.01,0.27501888275146485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.01,0.27724863052368165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.01,0.28825279235839846
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.01,0.2880441665649414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.01,0.3061011123657226
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.01,0.31793216705322264
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.01,0.34569473266601564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.01,0.3947654342651367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.01,0.459153938293457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.01,0.567760009765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.01,0.6801087951660156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.01,0.9133197021484374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.01,1.1460819244384766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.01,1.6152972412109374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.01,2.081921844482422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,2,1,power_law_1.01,2.5462355041503906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,2,1,power_law_1.2,1.6905087280273439
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.01,0.03815423965454102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.01,0.07642687797546387
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.01,0.048889598846435546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.01,0.061304321289062505
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.01,0.0910585594177246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.01,0.12840448379516603
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.01,0.1379923152923584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.01,0.14055999755859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.01,0.14515839576721193
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.01,0.14655296325683592
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.01,0.1520364761352539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.01,0.15527168273925782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.01,0.15807552337646485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.01,0.1650752067565918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.01,0.17416255950927734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.01,0.17859519958496092
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.01,0.1926412773132324
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,2,1,power_law_1.2,2.1048326110839843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.01,0.22173376083374025
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.01,0.24900415420532224
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.01,0.32055553436279294
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.01,0.3908940887451172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.01,0.5384352111816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.01,0.6832294464111328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.01,0.9855935668945313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.01,1.316378936767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,2,1,power_law_1.2,0.0616096019744873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,2,1,power_law_1.2,0.08165120124816895
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,2,1,power_law_1.2,0.11959744453430177
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,2,1,power_law_1.2,0.17453119277954102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,2,1,power_law_1.2,0.22001855850219726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,2,1,power_law_1.2,0.3461894226074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,2,1,power_law_1.2,0.35814144134521486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,2,1,power_law_1.2,0.3781798553466797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,2,1,power_law_1.2,0.39652862548828127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,2,1,power_law_1.2,0.3974803161621094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,2,1,power_law_1.2,0.41875518798828126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,2,1,power_law_1.2,0.4399372863769531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,2,1,power_law_1.2,0.4508127975463867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,2,1,power_law_1.2,0.48535934448242185
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,2,1,power_law_1.2,0.49985088348388673
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,2,1,power_law_1.2,0.5006425476074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,2,1,power_law_1.2,0.5138508987426758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,2,1,power_law_1.2,0.5812006378173828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,2,1,power_law_1.2,0.6268537521362305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,2,1,power_law_1.2,0.7657734680175781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,2,1,power_law_1.2,0.8996275329589845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,2,1,power_law_1.2,1.1951904296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.01,1.9952851867675783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,2,1,power_law_1.2,1.496820526123047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.01,2.6391482543945313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.01,3.2817037963867186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,2,1,power_law_1.2,2.0987429809570313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,2,1,power_law_1.2,2.705063781738281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,2,1,power_law_1.2,0.03858112096786499
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,2,1,power_law_1.2,0.06137663841247558
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,2,1,power_law_1.2,0.09183872222900391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,2,1,power_law_1.2,0.11154687881469727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,2,1,power_law_1.2,0.14836159706115723
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,2,1,power_law_1.2,0.22085887908935548
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,2,1,power_law_1.2,0.2282431983947754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,2,1,power_law_1.2,0.2348134422302246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,2,1,power_law_1.2,0.23903488159179687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,2,1,power_law_1.2,0.24028415679931642
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,2,1,power_law_1.2,0.24367040634155274
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,2,1,power_law_1.2,0.2728166389465332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,2,1,power_law_1.2,0.27511808395385745
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,2,1,power_law_1.2,0.28658048629760746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,2,1,power_law_1.2,0.2939423942565918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,2,1,power_law_1.2,0.2970457649230957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,2,1,power_law_1.2,0.30628927230834957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,2,1,power_law_1.2,0.3246688079833985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,2,1,power_law_1.2,0.3523756790161133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,2,1,power_law_1.2,0.40785343170166016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,2,1,power_law_1.2,0.4637164688110351
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,2,1,power_law_1.2,0.5734809494018555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,2,1,power_law_1.2,0.6905222320556641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,2,1,power_law_1.2,3.927108459472656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,2,1,power_law_1.2,0.9216146850585938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,2,1,power_law_1.2,1.164155502319336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,2,1,power_law_1.2,5.139714965820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,2,1,power_law_1.2,1.617472686767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,2,1,power_law_1.2,0.03789184093475342
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,2,1,power_law_1.2,0.04839104175567627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,2,1,power_law_1.2,0.060478720664978024
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,2,1,power_law_1.2,0.07174399852752686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,2,1,power_law_1.2,0.0857919979095459
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,2,1,power_law_1.2,0.13080512046813966
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,2,1,power_law_1.2,0.1385331153869629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,2,1,power_law_1.2,0.14371583938598634
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,2,1,power_law_1.2,0.14821632385253908
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,2,1,power_law_1.2,0.14838080406188964
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,2,1,power_law_1.2,0.15404159545898438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,2,1,power_law_1.2,0.15919936180114747
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,2,1,power_law_1.2,0.16270591735839843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,2,1,power_law_1.2,0.17327167510986327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,2,1,power_law_1.2,0.18287616729736328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,2,1,power_law_1.2,0.19045440673828123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,2,1,power_law_1.2,6.352391967773437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,2,1,power_law_1.2,0.20867008209228516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,2,1,power_law_1.2,2.0993753051757813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,2,1,power_law_1.2,0.2526688003540039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,2,1,power_law_1.2,0.2983910369873047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,2,1,power_law_1.2,0.40059326171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,2,1,power_law_1.2,0.5029913711547851
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,2,1,power_law_1.2,2.5520640563964845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,2,1,power_law_1.2,0.7228102111816407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,2,1,power_law_1.2,0.9535456085205078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,2,1,power_law_1.2,1.3887525939941407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,2,1,power_law_1.2,1.8248165893554689
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.01,0.09677760124206543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.01,0.15178751945495605
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.01,0.3045715141296387
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.01,0.4543833541870117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.01,0.6017478561401367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.01,0.9421651458740234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.01,0.9895552062988282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.01,1.0291487884521484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.01,1.0654905700683595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.01,1.0859078216552736
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.01,1.1391040039062499
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.01,1.1988076782226562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.01,1.2639405059814455
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.01,1.3093350219726563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.01,1.3234732055664062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.01,1.328616943359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.01,1.4206124877929687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.01,1.5684249877929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.01,1.6901158142089845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.01,1.966367950439453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,2,1,power_law_1.2,2.720258483886719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.01,2.3190106201171874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.01,2.971642761230469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.01,3.6605838012695315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.01,5.091318359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.01,6.527864990234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.01,9.367432861328124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.01,12.261185302734376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.01,0.06861567974090577
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.01,0.10990143775939942
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.01,0.17124992370605469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.01,0.24795200347900392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.01,0.3335232162475586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.01,0.5136134338378906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.01,0.5439603042602539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.01,0.5639353561401367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.01,0.5795590209960937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,2,1,power_law_1.01,15.123668212890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.01,0.5876985549926758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.01,0.6023884963989258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,2,1,power_law_1.2,3.617099609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.01,0.6161689758300781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.01,0.6665433502197265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.01,0.6937023925781249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.01,0.6883430480957031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.01,0.6975769805908203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.01,0.7209561920166015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.01,0.7680159759521484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.01,0.8007981109619141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.01,0.8830188751220703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.01,1.0045049285888672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.01,1.2092511749267578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.01,1.4225241088867189
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.01,1.8659423828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.01,2.3095091247558592
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.01,3.2373965454101565
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.01,4.076061401367188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,2,1,power_law_1.01,4.987675476074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.01,0.061959037780761714
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.01,0.15104512214660643
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.01,0.07797247886657714
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.01,0.19792383193969726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.01,0.31087551116943357
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.01,0.3287065505981445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.01,0.10644096374511718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,2,1,power_law_1.2,4.472777709960938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.01,0.34142337799072264
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.01,0.35352638244628903
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.01,0.35795902252197265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.01,0.37920513153076174
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.01,0.3828166580200195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.01,0.4026752090454101
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.01,0.4090937423706055
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.01,0.41129089355468745
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.01,0.3689548873901367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.01,0.4391519927978516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.01,0.4903404617309571
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.01,0.5319551849365235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.01,0.6303705596923829
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.01,0.750224609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.01,0.9922195434570312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.01,1.2349318695068359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.01,1.7371743774414063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.01,2.2645523071289064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,2,1,power_law_1.2,0.0970195198059082
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,2,1,power_law_1.2,0.15154239654541016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,2,1,power_law_1.2,0.2963065528869629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,2,1,power_law_1.2,0.4017619323730469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,2,1,power_law_1.2,0.5360192108154297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.01,3.319810485839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,2,1,power_law_1.2,0.9309158325195312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,2,1,power_law_1.2,0.9746975708007813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.01,4.451377258300782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,2,1,power_law_1.2,1.0353561401367188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,2,1,power_law_1.2,1.08189697265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,2,1,power_law_1.2,1.1174790191650392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,2,1,power_law_1.2,1.1760063934326173
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,2,1,power_law_1.2,1.2357965087890626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,2,1,power_law_1.2,1.306826171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,2,1,power_law_1.2,1.31943359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,2,1,power_law_1.2,1.3775334167480469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,2,1,power_law_1.2,1.4134278869628907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,2,1,power_law_1.2,1.4547468566894532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,2,1,power_law_1.2,1.5858546447753905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,2,1,power_law_1.2,1.7138418579101562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,2,1,power_law_1.2,2.0321688842773438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,2,1,power_law_1.2,2.3548748779296877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,2,1,power_law_1.2,3.052367858886719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,2,1,power_law_1.2,3.7278900146484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,2,1,power_law_1.2,5.15673095703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,2,1,power_law_1.2,6.578115234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,2,1,power_law_1.01,5.537091674804687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,2,1,power_law_1.2,9.444231567382811
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,2,1,power_law_1.2,0.07393599987030029
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,2,1,power_law_1.2,0.109552001953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,2,1,power_law_1.2,0.22967872619628907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,2,1,power_law_1.2,0.16455488204956054
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,2,1,power_law_1.2,0.31310592651367186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,2,1,power_law_1.2,0.5157516860961915
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,2,1,power_law_1.2,0.5384447860717774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,2,1,power_law_1.2,0.5659711837768555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,2,1,power_law_1.2,0.5876160049438477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,2,1,power_law_1.2,0.5993945693969727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,2,1,power_law_1.2,0.6158316802978516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,2,1,power_law_1.2,0.6244128036499024
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,2,1,power_law_1.2,0.6898553466796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,2,1,power_law_1.2,0.696600341796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,2,1,power_law_1.2,0.7150220489501953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,2,1,power_law_1.2,0.7249043273925782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,2,1,power_law_1.2,0.7449017333984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,2,1,power_law_1.2,0.7758975982666015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,2,1,power_law_1.2,0.8182975769042968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,2,1,power_law_1.2,0.9136998748779297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,2,1,power_law_1.2,1.01793212890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,2,1,power_law_1.2,1.2378809356689453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,2,1,power_law_1.2,1.4457254028320312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,2,1,power_law_1.2,1.900853729248047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,2,1,power_law_1.2,2.335333099365234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,2,1,power_law_1.2,12.343853759765626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,2,1,power_law_1.2,3.225696716308594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,2,1,power_law_1.2,0.06105984210968017
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,2,1,power_law_1.2,0.07455039978027343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,2,1,power_law_1.2,0.10297023773193359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,2,1,power_law_1.2,0.13786304473876954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,2,1,power_law_1.2,0.18354047775268553
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,2,1,power_law_1.2,0.30619007110595703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,2,1,power_law_1.2,0.32520065307617185
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,2,1,power_law_1.2,0.3459052658081055
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,2,1,power_law_1.2,15.18218994140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,2,1,power_law_1.2,0.3586297607421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,2,1,power_law_1.2,0.36864768981933593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,2,1,power_law_1.2,0.3822060775756836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,2,1,power_law_1.2,4.1216909790039065
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,2,1,power_law_1.2,0.38833473205566404
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,2,1,power_law_1.2,0.40188350677490237
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,2,1,power_law_1.2,0.4126015853881836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,2,1,power_law_1.2,5.036589965820313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,2,1,power_law_1.2,0.5395558547973633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,2,1,power_law_1.2,0.4336985778808594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,2,1,power_law_1.2,0.44619583129882817
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,2,1,power_law_1.2,0.47760959625244137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,2,1,power_law_1.2,0.6108038330078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,2,1,power_law_1.2,0.7744985961914063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,2,1,power_law_1.2,0.94957763671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,2,1,power_law_1.2,1.3338272094726562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,2,1,power_law_1.2,1.7409747314453123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,2,1,power_law_1.2,2.5861767578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,2,1,power_law_1.2,3.4485620117187503
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.01,0.10081088066101074
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.01,0.1419961643218994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.01,0.2944473648071289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.01,0.5111135864257813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.01,0.8489497375488281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.01,1.1770182037353516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.01,1.3439610290527344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,2,1,power_law_1.2,5.214203491210937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.01,1.7257522583007812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.01,1.8129344177246094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.01,1.8930496215820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.01,1.9549197387695312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.01,1.984003143310547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.01,2.0747827148437503
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.01,2.1871731567382815
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.01,2.2838893127441406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.01,2.442923583984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.01,2.460715484619141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.01,2.6004965209960935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.01,2.766810302734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.01,3.0605569458007813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.01,3.344453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.01,3.9195333862304684
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.01,4.549906005859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.01,5.785783081054687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,2,1,power_law_1.2,7.003782958984376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.01,7.0682373046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.01,0.06989888191223145
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.01,0.0958124828338623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.01,0.15736512184143064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.01,0.27738176345825194
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.01,0.46767040252685543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.01,0.6548185729980469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.01,0.7723219299316406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.01,0.9653472137451171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.01,12.379422607421876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.01,1.0093119812011717
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.01,1.049837417602539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,2,1,power_law_1.01,15.138221435546876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.01,1.0840499114990234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.01,1.0990541076660156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.01,1.1686029052734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.01,1.1355935668945312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.01,1.1795187377929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.01,1.220791015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.01,1.3104876708984374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.01,1.3595628356933593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.01,1.4112345886230468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.01,1.5149618530273439
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.01,1.6152485656738282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.01,1.811927032470703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.01,2.030812225341797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.01,2.4702272033691406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.01,2.8888632202148434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.01,3.8158688354492183
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.01,4.67246337890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,2,1,power_law_1.01,5.5582385253906255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.01,0.08333120346069336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.01,0.08564288139343261
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.01,0.08774144172668456
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.01,0.09508607864379884
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.01,0.0857203197479248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.01,0.10635135650634767
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.01,0.117575044631958
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.01,0.12721856117248537
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.01,0.14002367973327637
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.01,0.15041600227355958
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.01,0.1710188865661621
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.01,0.20581567764282224
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.01,0.22686016082763674
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.01,0.27087039947509767
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.01,0.3262227249145508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.01,0.373070068359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.01,0.47494014739990237
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.01,0.6810233306884765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.01,0.884703369140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.01,1.2930738830566406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.01,1.689949493408203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.01,2.4913133239746097
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.01,3.3072824096679687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,2,1,power_law_1.2,0.10075712203979492
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,2,1,power_law_1.2,0.14272255897521974
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,2,1,power_law_1.2,0.2839097595214844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,2,1,power_law_1.2,0.4911199951171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,2,1,power_law_1.2,0.7572889709472657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,2,1,power_law_1.2,1.0677120208740234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,2,1,power_law_1.2,1.2345804595947265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,2,1,power_law_1.2,1.6814930725097657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,2,1,power_law_1.2,1.7584538269042969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.01,4.941737060546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,2,1,power_law_1.2,1.8157061767578124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,2,1,power_law_1.2,1.890045471191406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,2,1,power_law_1.2,1.9850079345703127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,2,1,power_law_1.2,2.0612908935546876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,2,1,power_law_1.2,2.1869964599609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,2,1,power_law_1.2,2.329811859130859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,2,1,power_law_1.2,2.469640350341797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,2,1,power_law_1.2,2.4995423889160158
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,2,1,power_law_1.2,2.678164367675781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,2,1,power_law_1.2,2.810582275390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,2,1,power_law_1.2,3.1320761108398436
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,2,1,power_law_1.2,3.46873779296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,2,1,power_law_1.2,4.069022216796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,2,1,power_law_1.2,4.710216979980468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,2,1,power_law_1.2,5.961631469726562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,2,1,power_law_1.2,7.308922729492187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,2,1,power_law_1.2,9.9470751953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,2,1,power_law_1.2,0.07012608051300048
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,2,1,power_law_1.2,0.09623488426208496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,2,1,power_law_1.2,0.15827839851379394
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,2,1,power_law_1.2,0.27919424057006836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,2,1,power_law_1.2,12.721485595703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,2,1,power_law_1.2,0.4296185684204102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,2,1,power_law_1.2,0.6151494216918946
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,2,1,power_law_1.2,0.7181094360351563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,2,1,power_law_1.2,0.9465261077880859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,2,1,power_law_1.2,0.9846182250976563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,2,1,power_law_1.2,1.0150860595703126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,2,1,power_law_1.2,1.0490003204345704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,2,1,power_law_1.2,1.0929350280761718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,2,1,power_law_1.2,1.1208076477050781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,2,1,power_law_1.2,1.1551776123046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,2,1,power_law_1.2,1.1808780670166015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,2,1,power_law_1.2,1.2104704284667969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,2,1,power_law_1.2,1.3174899291992186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,2,1,power_law_1.2,1.3872845458984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,2,1,power_law_1.2,1.5577043151855468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,2,1,power_law_1.2,15.3620751953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,2,1,power_law_1.2,1.684767303466797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,2,1,power_law_1.2,1.9042906188964843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,2,1,power_law_1.2,2.1242662048339844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,2,1,power_law_1.2,1.4357830810546877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,2,1,power_law_1.2,2.5696896362304686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,2,1,power_law_1.2,3.006402587890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,2,1,power_law_1.2,3.8914349365234373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,2,1,power_law_1.2,0.08513088226318359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,2,1,power_law_1.2,4.781493225097656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,2,1,power_law_1.2,0.0863526439666748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,2,1,power_law_1.2,5.677277832031249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,2,1,power_law_1.2,0.09497344017028808
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,2,1,power_law_1.2,0.11694272041320801
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,2,1,power_law_1.2,0.12821056365966796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,2,1,power_law_1.2,0.1391865634918213
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,2,1,power_law_1.2,0.15035327911376953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,2,1,power_law_1.2,0.17281919479370117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,2,1,power_law_1.2,0.20548543930053711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,2,1,power_law_1.2,0.22802560806274413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,2,1,power_law_1.2,0.27290431976318363
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,2,1,power_law_1.2,0.0854860782623291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,2,1,power_law_1.2,0.3286163330078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,2,1,power_law_1.2,0.08619327545166015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,2,1,power_law_1.2,0.10466624259948731
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,2,1,power_law_1.2,0.8849369812011719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,2,1,power_law_1.2,0.3708812713623047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,2,1,power_law_1.2,0.48230976104736334
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,2,1,power_law_1.2,0.6811129760742187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,2,1,power_law_1.2,1.2883949279785156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,2,1,power_law_1.2,1.688983612060547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,2,1,power_law_1.2,2.510899200439453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,2,1,power_law_1.2,3.2970062255859376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,2,1,power_law_1.2,4.921939697265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.01,0.10868224143981933
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.01,0.1438335990905762
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.01,0.17092479705810545
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.01,0.25764863967895507
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.01,0.27603647232055667
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.01,0.2919615936279297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.01,0.29973951339721683
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.01,0.304147834777832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.01,0.3129439926147461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.01,0.35112064361572265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.01,0.36376190185546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.01,0.37129150390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.01,0.3709145736694336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.01,0.3729292678833008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.01,0.3891462326049805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.01,0.42564224243164067
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.01,0.4570316696166993
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.01,0.5281273651123046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.01,0.6111142349243164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.01,0.7779468536376953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.01,0.9635923004150391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.01,1.3255993652343752
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.01,1.692126007080078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.01,2.4229452514648435
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.01,3.163172607421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,2,1,power_law_1.01,3.8964447021484374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,power_law_1.01,0.0377017593383789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,power_law_1.01,0.053864960670471196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,power_law_1.01,0.0668345594406128
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,power_law_1.01,0.07927040100097657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,power_law_1.01,0.10731712341308594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,power_law_1.01,0.11112128257751466
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,power_law_1.01,0.11597311973571778
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,power_law_1.01,0.12199104309082032
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,power_law_1.01,0.12459391593933104
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,power_law_1.01,0.12758848190307617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,power_law_1.01,0.1346841621398926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,power_law_1.01,0.1382431983947754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,power_law_1.01,0.04181503772735596
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,power_law_1.01,0.14138943672180176
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,power_law_1.01,0.11841600418090821
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,power_law_1.01,0.14972415924072266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,power_law_1.01,0.1686777687072754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,power_law_1.01,0.12834303855895995
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,power_law_1.01,0.18662527084350586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,power_law_1.01,0.22603456497192384
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,power_law_1.01,0.27133119583129883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,power_law_1.01,0.35664703369140627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,power_law_1.01,0.453138542175293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,power_law_1.01,0.647325439453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,power_law_1.01,0.8439558410644532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,power_law_1.01,1.2365708923339844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,2,1,power_law_1.2,0.04329984188079834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,2,1,power_law_1.2,0.08557696342468261
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,2,1,power_law_1.2,0.1284838390350342
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,2,1,power_law_1.2,0.16270719528198244
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,2,1,power_law_1.2,0.25701375961303713
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,2,1,power_law_1.2,0.27985727310180664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,2,1,power_law_1.2,0.29496576309204103
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,2,1,power_law_1.2,0.30063743591308595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,2,1,power_law_1.2,0.05800704002380371
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,2,1,power_law_1.2,0.3059321594238281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,2,1,power_law_1.2,0.3649958419799805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,2,1,power_law_1.2,0.36975166320800784
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,2,1,power_law_1.2,0.3759123229980469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,2,1,power_law_1.2,0.3235116958618164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,2,1,power_law_1.2,0.3762809753417969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,2,1,power_law_1.2,0.38259201049804686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,2,1,power_law_1.2,0.4012377548217773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,2,1,power_law_1.2,0.4621696090698242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,2,1,power_law_1.2,0.43326080322265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,2,1,power_law_1.2,0.5428134536743163
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,2,1,power_law_1.2,0.6246847915649414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,2,1,power_law_1.2,0.7987506866455079
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,2,1,power_law_1.2,0.9773612976074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,power_law_1.01,1.6688435363769532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,2,1,power_law_1.2,1.3394355773925781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,2,1,power_law_1.01,2.110352020263672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,2,1,power_law_1.2,1.7100575256347657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2688,1856,6,128,2,1,power_law_1.2,0.03776511907577514
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2688,1856,6,128,2,1,power_law_1.2,0.04442624092102051
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2688,1856,6,128,2,1,power_law_1.2,0.05352447986602783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2688,1856,6,128,2,1,power_law_1.2,0.06298304080963135
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2688,1856,6,128,2,1,power_law_1.2,0.07638400077819825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2688,1856,6,128,2,1,power_law_1.2,0.10638719558715821
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,2,1,power_law_1.2,2.4436685180664064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2688,1856,6,128,2,1,power_law_1.2,0.11187904357910157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2688,1856,6,128,2,1,power_law_1.2,0.11814144134521484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2688,1856,6,128,2,1,power_law_1.2,0.12008064270019532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2688,1856,6,128,2,1,power_law_1.2,0.12355008125305175
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2688,1856,6,128,2,1,power_law_1.2,0.12571200370788574
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2688,1856,6,128,2,1,power_law_1.2,0.13186623573303222
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2688,1856,6,128,2,1,power_law_1.2,0.13395520210266115
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2688,1856,6,128,2,1,power_law_1.2,0.14093440055847167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,2,1,power_law_1.2,3.1808587646484376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2688,1856,6,128,2,1,power_law_1.2,0.16342912673950197
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,2,1,power_law_1.2,3.91690185546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2688,1856,6,128,2,1,power_law_1.2,0.1464883232116699
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2688,1856,6,128,2,1,power_law_1.2,0.15195775985717774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2688,1856,6,128,2,1,power_law_1.2,0.18828287124633789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2688,1856,6,128,2,1,power_law_1.2,0.21517696380615234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2688,1856,6,128,2,1,power_law_1.2,0.4896543884277344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2688,1856,6,128,2,1,power_law_1.2,0.27964096069335936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2688,1856,6,128,2,1,power_law_1.2,0.34510143280029293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2688,1856,6,128,2,1,power_law_1.2,0.6425452423095703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2688,1856,6,128,2,1,power_law_1.2,0.9491289520263673
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2688,1856,6,128,2,1,power_law_1.2,1.2816120910644533
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.01,0.17286527633666993
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2688,1856,6,128,2,1,power_law_1.2,1.9600473022460938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.01,0.2750252723693848
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.01,0.8210566711425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.01,1.1466636657714844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.01,0.4843993759155273
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.01,1.8783833312988283
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.01,2.0932109069824216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.01,2.0029779052734376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.01,2.1597669982910155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2688,1856,6,128,2,1,power_law_1.2,2.6478195190429688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.01,2.1962815856933595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.01,2.3012646484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.01,2.545726776123047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.01,2.4194848632812502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.01,2.8877966308593748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.01,3.174105529785156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.01,3.5323992919921876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.01,4.20682373046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.01,3.13465087890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.01,3.391706237792969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.01,3.940640563964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.01,4.531112976074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.01,5.806663818359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2688,1856,6,128,2,1,power_law_1.2,3.2806201171875005
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.01,7.1547961425781255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.01,9.9238525390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.01,12.66573974609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.01,0.10848575592041017
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.01,18.231368408203124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.01,0.17376127243041992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.01,0.2836275291442871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.01,23.97525634765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.01,0.46599166870117187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.01,0.6502796936035156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.01,1.044013442993164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.01,1.2032377624511719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.01,1.1133657836914064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,2,1,power_law_1.01,29.93702392578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.01,1.2202751922607422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.01,1.161829147338867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.01,1.2469312286376952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.01,1.3045516967773438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.01,1.3556723022460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.01,1.3564524841308594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.01,1.4059327697753905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.01,1.28457275390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.01,1.4570118713378908
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.01,1.6227296447753907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.01,1.7145184326171876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.01,1.917606964111328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.01,2.1360287475585937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.01,2.6013522338867188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.01,3.0565151977539062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.01,3.9836523437499998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.01,4.947820739746094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.01,6.9020800781250005
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.01,8.858795166015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,2,1,power_law_1.01,10.772021484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,power_law_1.01,0.12498623847961425
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,power_law_1.01,0.1316096019744873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,power_law_1.01,0.12534015655517577
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,power_law_1.01,0.14180288314819336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,power_law_1.01,0.12538304328918456
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,power_law_1.01,0.15018688201904298
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,power_law_1.01,0.1604947280883789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,power_law_1.01,0.16855424880981446
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,power_law_1.01,0.17773632049560545
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,power_law_1.01,0.19683263778686524
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,power_law_1.01,0.12680895805358888
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,power_law_1.01,0.23346687316894532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,power_law_1.01,0.2512230491638184
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,power_law_1.01,0.28789247512817384
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,power_law_1.01,0.34325313568115234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,power_law_1.01,0.381710090637207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,power_law_1.01,0.4753100967407226
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,power_law_1.01,0.6640902709960937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.01,0.8541375732421874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.01,1.2377632141113282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.01,1.609385528564453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.01,2.371435546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.01,3.1199691772460936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.01,4.662662963867187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,2,1,power_law_1.2,0.17301824569702148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.01,6.157915649414063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,2,1,power_law_1.2,0.2743929672241211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,2,1,power_law_1.2,0.4870502471923828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,2,1,power_law_1.2,0.7531731414794922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,2,1,power_law_1.2,1.8432684326171873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,2,1,power_law_1.2,1.0296876525878906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,2,1,power_law_1.2,2.0012300109863284
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,2,1,power_law_1.2,2.067659454345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,2,1,power_law_1.2,2.161049652099609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,2,1,power_law_1.2,2.2469215393066406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,2,1,power_law_1.2,2.3589593505859376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,2,1,power_law_1.2,2.5141734313964843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,2,1,power_law_1.2,2.6610983276367186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,2,1,power_law_1.2,3.003908996582031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,2,1,power_law_1.2,3.344599609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,2,1,power_law_1.2,3.66444091796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,2,1,power_law_1.2,4.340064697265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,2,1,power_law_1.2,3.256222229003906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,2,1,power_law_1.2,3.499695434570312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,2,1,power_law_1.2,4.037683715820313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,2,1,power_law_1.2,4.656526794433594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,2,1,power_law_1.2,5.9402398681640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,2,1,power_law_1.2,7.337892456054687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,2,1,power_law_1.2,10.049817504882814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,2,1,power_law_1.2,12.85673583984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,2,1,power_law_1.2,0.10540032386779785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,2,1,power_law_1.2,18.9447314453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,2,1,power_law_1.2,0.28584512710571286
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,2,1,power_law_1.2,0.4314879989624023
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,2,1,power_law_1.2,0.5996192169189454
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,2,1,power_law_1.2,0.16753599166870117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,2,1,power_law_1.2,1.0324569702148438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,2,1,power_law_1.2,1.107943649291992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,2,1,power_law_1.2,1.1482195281982421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,2,1,power_law_1.2,1.1955270385742187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,2,1,power_law_1.2,1.2315046691894531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,2,1,power_law_1.2,1.2709951782226563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,2,1,power_law_1.2,1.3060345458984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,2,1,power_law_1.2,1.3333094787597655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,2,1,power_law_1.2,1.3951589965820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,2,1,power_law_1.2,1.41382080078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,2,1,power_law_1.2,1.4468409729003906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,2,1,power_law_1.2,1.5139552307128905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,2,1,power_law_1.2,1.6932243347167968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,2,1,power_law_1.2,1.7828012084960938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,2,1,power_law_1.2,1.9940985107421874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,2,1,power_law_1.2,2.2064031982421874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,2,1,power_law_1.2,2.6670324707031248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,2,1,power_law_1.2,3.1114578247070312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,2,1,power_law_1.2,4.074583129882813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,2,1,power_law_1.2,24.743791503906248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,2,1,power_law_1.2,5.0304846191406245
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,2,1,power_law_1.2,6.973290405273437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,2,1,power_law_1.2,0.12559871673583983
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,2,1,power_law_1.2,0.12554880142211916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,2,1,power_law_1.2,0.1258886432647705
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,2,1,power_law_1.2,0.12669504165649415
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,2,1,power_law_1.2,0.13158016204833983
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,2,1,power_law_1.2,0.14110912322998045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,2,1,power_law_1.2,0.15084799766540527
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,2,1,power_law_1.2,0.1599180793762207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,2,1,power_law_1.2,8.941992797851563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,2,1,power_law_1.2,0.16923328399658205
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,2,1,power_law_1.2,0.17822912216186523
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,2,1,power_law_1.2,31.839189453124998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,2,1,power_law_1.2,0.19617279052734377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,2,1,power_law_1.2,0.23278207778930665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,2,1,power_law_1.2,0.28728960037231444
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,2,1,power_law_1.2,0.344318733215332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,2,1,power_law_1.2,10.839194335937501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,2,1,power_law_1.2,0.24962175369262693
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,2,1,power_law_1.2,0.38010814666748044
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,2,1,power_law_1.2,0.47706878662109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,2,1,power_law_1.2,1.238260498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,2,1,power_law_1.2,1.610216369628906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,2,1,power_law_1.2,0.6622470092773438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,2,1,power_law_1.2,0.8541395568847656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,2,1,power_law_1.2,2.3784947204589844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,2,1,power_law_1.2,3.1227801513671873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,2,1,power_law_1.2,4.636260986328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,2,1,power_law_1.2,6.15024658203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.01,0.10709568023681641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.01,0.10254912376403809
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.01,0.13846783638000487
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.01,0.1687763214111328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.01,0.17001472473144533
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.01,0.1702067184448242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.01,0.16954879760742186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.01,0.17048063278198242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.01,0.1810054397583008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.01,0.21539327621459964
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.01,0.215677433013916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.01,0.22316095352172854
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.01,0.23440832138061524
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.01,0.2411257553100586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.01,0.27534784317016603
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.01,0.32549182891845707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.01,0.3913024139404297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.01,0.46694976806640626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.01,0.6285433578491211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.01,0.7867603302001953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.01,1.1154239654541016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.01,1.4431251525878905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.01,2.103422088623047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.01,2.7591641235351565
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.01,5.397479858398437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.01,0.061686401367187504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.01,0.06753407955169678
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.01,0.09025216102600098
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.01,0.09619839668273926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.01,0.09792063713073731
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.01,0.09864447593688966
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.01,0.09914496421813965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.01,0.09949312210083008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.01,0.10153152465820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.01,0.10383423805236816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.01,0.1039846420288086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.01,0.045735039710998536
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.01,0.1060524845123291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.01,0.11006784439086914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.01,0.1223731231689453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.01,0.12629823684692382
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.01,0.1329100799560547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.01,0.15285375595092773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.01,0.17763711929321288
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.01,0.2278156852722168
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.01,0.2646976089477539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.01,0.3647974395751953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.01,0.46481601715087895
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.01,0.6585075378417968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.01,0.8484544372558593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,4,1,power_law_1.01,7.011481323242188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.01,1.2374700927734374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.01,1.614207305908203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.01,0.02534271955490112
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.01,0.05109824180603027
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.01,0.051742720603942874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.01,0.05225279808044434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.01,0.052248959541320805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.01,0.05195648193359374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.01,0.05175424098968506
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.01,0.05234816074371338
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.01,0.05248576164245605
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,4,1,power_law_1.01,2.009205169677734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.01,0.033272318840026856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.01,0.039717121124267576
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.01,0.053119359016418455
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.01,0.053811202049255366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.01,0.05802303791046143
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.01,0.06131264209747315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.01,0.0646227216720581
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.01,0.06914495944976806
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.01,0.0818278408050537
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.01,0.09788479804992675
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.01,0.12537792205810547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.01,0.1543065643310547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.01,0.2101055908203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.01,0.267390079498291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.01,0.3826502227783203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.01,0.5000889587402344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,4,1,power_law_1.2,0.07558207988739014
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,4,1,power_law_1.2,0.09078271865844727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,4,1,power_law_1.2,0.09720512390136718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,4,1,power_law_1.2,0.13917247772216795
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,4,1,power_law_1.2,0.1646348762512207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,4,1,power_law_1.2,0.17049407958984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,4,1,power_law_1.2,0.17124160766601562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,4,1,power_law_1.2,0.1713555145263672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,4,1,power_law_1.2,0.17363712310791016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,4,1,power_law_1.2,0.18048704147338868
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,4,1,power_law_1.2,0.21623104095458984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,4,1,power_law_1.2,0.21999935150146482
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,4,1,power_law_1.2,0.222608642578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,4,1,power_law_1.2,0.23531455993652345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,4,1,power_law_1.2,0.24292736053466796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,4,1,power_law_1.2,0.28853504180908207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,4,1,power_law_1.2,0.3259072113037109
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,4,1,power_law_1.2,0.39305023193359373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.01,0.7275929260253906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,4,1,power_law_1.2,0.4721395111083984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,4,1,power_law_1.2,0.6313017654418945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,4,1,power_law_1.2,0.7895776367187499
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,4,1,power_law_1.2,1.1170252990722656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.01,0.9559686279296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,4,1,power_law_1.2,1.4412832641601563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,4,1,power_law_1.2,2.0985752868652345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.01,1.1976595306396485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,4,1,power_law_1.2,2.752227783203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,4,1,power_law_1.2,0.05067647933959961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,4,1,power_law_1.2,0.06020607948303223
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,4,1,power_law_1.2,0.063439359664917
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,4,1,power_law_1.2,0.08982912063598633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,4,1,power_law_1.2,0.09321599960327148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,4,1,power_law_1.2,0.096178560256958
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,4,1,power_law_1.2,0.09916671752929687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,4,1,power_law_1.2,0.0990719985961914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,4,1,power_law_1.2,0.10088512420654297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,4,1,power_law_1.2,0.10133440017700195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,4,1,power_law_1.2,0.103121919631958
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,4,1,power_law_1.2,0.10430080413818359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,4,1,power_law_1.2,0.10399935722351075
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,4,1,power_law_1.2,0.11497471809387208
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,4,1,power_law_1.2,0.12546624183654784
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,4,1,power_law_1.2,0.13536895751953124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,4,1,power_law_1.2,0.13241344451904297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,4,1,power_law_1.2,0.15131327629089356
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,4,1,power_law_1.2,0.17736255645751953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,4,1,power_law_1.2,0.21890687942504883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,4,1,power_law_1.2,0.2652332878112793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,4,1,power_law_1.2,4.081089782714844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,4,1,power_law_1.2,0.3652582550048828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,4,1,power_law_1.2,0.4704985427856445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,4,1,power_law_1.2,0.658703384399414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,4,1,power_law_1.2,5.40381591796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,4,1,power_law_1.2,0.8575910186767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,4,1,power_law_1.2,7.268828735351562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,4,1,power_law_1.2,0.02554879903793335
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,4,1,power_law_1.2,0.03337599992752075
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,4,1,power_law_1.2,0.036038401126861574
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,4,1,power_law_1.2,0.05115263938903809
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,4,1,power_law_1.2,0.0508896017074585
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,4,1,power_law_1.2,0.05205440044403076
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,4,1,power_law_1.2,0.05219967842102051
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,4,1,power_law_1.2,0.051476478576660156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,4,1,power_law_1.2,0.052284159660339356
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,4,1,power_law_1.2,0.052344322204589844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,4,1,power_law_1.2,0.05254591941833496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,4,1,power_law_1.2,0.053477120399475095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,4,1,power_law_1.2,0.05484288215637208
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,4,1,power_law_1.2,0.05931583881378174
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,4,1,power_law_1.2,0.06223552227020264
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,4,1,power_law_1.2,0.06632575988769532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,4,1,power_law_1.2,1.2328179168701172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,4,1,power_law_1.2,1.6220710754394532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,4,1,power_law_1.2,0.06939712047576904
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,4,1,power_law_1.2,0.12746687889099123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,4,1,power_law_1.2,0.08523712158203126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,4,1,power_law_1.2,0.09944255828857422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,4,1,power_law_1.2,0.1569478416442871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,4,1,power_law_1.2,2.0149778747558593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,4,1,power_law_1.2,0.21743295669555662
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,4,1,power_law_1.2,0.3944460678100586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,4,1,power_law_1.2,0.27361536026000977
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,4,1,power_law_1.2,0.514582405090332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.01,0.09989567756652831
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.01,0.1283091163635254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.01,0.15060607910156248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.01,0.23824384689331052
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.01,0.2501907157897949
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.01,0.252740478515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.01,0.2546636772155762
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.01,0.2509939193725586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.01,0.2565548706054687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.01,0.27699520111083986
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.01,0.33097152709960936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.01,0.35104190826416015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.01,0.36409793853759764
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.01,0.377448959350586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.01,0.3952435302734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.01,0.42823486328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.01,0.4952057647705078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.01,0.6288332748413086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.01,0.7581247711181641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.01,1.032111358642578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.01,1.3068428039550781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.01,1.871418914794922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.01,2.4338777160644534
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.01,3.5660491943359376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,4,1,power_law_1.2,0.7460006713867188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.01,4.690244445800781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,4,1,power_law_1.2,0.9938662719726562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.01,6.94121826171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.01,0.06148863792419433
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.01,0.0799398422241211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.01,0.09921983718872071
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.01,0.13374143600463867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.01,0.1432767963409424
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.01,0.1448953628540039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.01,0.14641152381896974
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.01,0.14551551818847658
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.01,0.14681920051574707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.01,0.14854592323303223
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.01,0.15015616416931152
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.01,0.15074239730834962
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.01,0.15383551597595216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.01,0.1667424011230469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.01,0.1714240074157715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.01,0.1737708854675293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.01,0.1834592056274414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.01,0.2175507164001465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.01,0.25737087249755863
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.01,0.328210563659668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.01,0.38861438751220706
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.01,0.5392544174194336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.01,9.455572509765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,4,1,power_law_1.2,1.224865264892578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.01,0.6802425384521484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.01,0.9811135864257812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.01,1.2570668792724609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,4,1,power_law_1.01,13.278778076171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.01,1.8328601074218749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.01,2.4855609130859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.01,0.03113663911819458
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.01,0.07590144157409669
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.01,0.07755199909210206
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.01,0.07797376155853272
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.01,0.07787327766418457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.01,0.07671807765960693
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.01,0.07663360118865967
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.01,0.07852799892425537
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.01,0.07883647918701173
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,4,1,power_law_1.01,3.022443542480469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.01,0.04759232044219971
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.01,0.056076798439025875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.01,0.07864384174346924
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.01,0.08077183723449707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.01,0.08603391647338868
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.01,0.08849791526794434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.01,0.08963839530944824
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.01,0.09840895652770995
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.01,0.12255104064941406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.01,0.14329536437988283
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.01,0.186180477142334
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.01,0.22634815216064452
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.01,0.3120639991760254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.01,0.4021785736083984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.01,0.573372802734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.01,0.7420333099365235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,4,1,power_law_1.2,0.09920319557189941
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,4,1,power_law_1.2,0.12391872406005859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,4,1,power_law_1.2,0.13614144325256347
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,4,1,power_law_1.2,0.24214784622192384
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,4,1,power_law_1.2,0.23846080780029294
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,4,1,power_law_1.2,0.24979711532592774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,4,1,power_law_1.2,0.25294527053833005
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,4,1,power_law_1.2,0.2481920051574707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,4,1,power_law_1.2,0.2599398422241211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,4,1,power_law_1.2,0.2767238426208496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,4,1,power_law_1.2,0.3384172821044922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,4,1,power_law_1.2,0.354205436706543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,4,1,power_law_1.2,0.35798336029052735
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,4,1,power_law_1.2,0.3813158416748047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,4,1,power_law_1.2,0.4012179183959961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,4,1,power_law_1.2,0.4383980941772461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,4,1,power_law_1.2,0.4897094345092773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,4,1,power_law_1.2,0.6297036743164062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.01,1.0968748474121095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,4,1,power_law_1.2,0.7638400268554688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,4,1,power_law_1.2,1.0312249755859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,4,1,power_law_1.2,1.31025146484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,4,1,power_law_1.2,1.8685824584960937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,4,1,power_law_1.2,2.424788513183594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.01,1.4664434814453124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,4,1,power_law_1.2,3.5522171020507813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,4,1,power_law_1.2,4.672119140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.01,1.8215295410156251
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,4,1,power_law_1.2,0.060399360656738274
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,4,1,power_law_1.2,0.07658368110656738
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,4,1,power_law_1.2,0.08510656356811523
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,4,1,power_law_1.2,0.13397695541381835
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,4,1,power_law_1.2,0.14095680236816405
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,4,1,power_law_1.2,0.144016637802124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,4,1,power_law_1.2,0.14545215606689452
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,4,1,power_law_1.2,0.14529664039611817
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,4,1,power_law_1.2,0.146693115234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,4,1,power_law_1.2,0.14725055694580078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,4,1,power_law_1.2,0.14879551887512207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,4,1,power_law_1.2,0.15134464263916017
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,4,1,power_law_1.2,0.15133824348449706
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,4,1,power_law_1.2,0.16590015411376952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,4,1,power_law_1.2,0.17568511962890626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,4,1,power_law_1.2,0.17745023727416992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,4,1,power_law_1.2,0.18870975494384765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,4,1,power_law_1.2,0.21725183486938476
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,4,1,power_law_1.2,0.25833152770996093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,4,1,power_law_1.2,0.3182579231262207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,4,1,power_law_1.2,0.39123584747314455
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,4,1,power_law_1.2,6.9464013671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,4,1,power_law_1.2,0.5409120178222656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,4,1,power_law_1.2,0.6890406036376954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,4,1,power_law_1.2,0.9720588684082031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,4,1,power_law_1.2,9.689899291992187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,4,1,power_law_1.2,1.2798572540283204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,4,1,power_law_1.2,13.97806884765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,4,1,power_law_1.2,1.8348089599609374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,4,1,power_law_1.2,0.045734400749206545
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,4,1,power_law_1.2,0.05056384086608887
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,4,1,power_law_1.2,0.07590720176696777
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,4,1,power_law_1.2,0.07459712028503418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,4,1,power_law_1.2,0.07803391933441162
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,4,1,power_law_1.2,0.07811327934265136
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,4,1,power_law_1.2,0.0766380786895752
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,4,1,power_law_1.2,0.07818751811981202
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,4,1,power_law_1.2,0.0786406421661377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,4,1,power_law_1.2,0.07928448200225831
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,4,1,power_law_1.2,0.08012224197387695
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,4,1,power_law_1.2,0.0799174404144287
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,4,1,power_law_1.2,0.08585984230041503
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,4,1,power_law_1.2,0.09058879852294922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,4,1,power_law_1.2,0.09275199890136718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,4,1,power_law_1.2,0.03118335962295532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,4,1,power_law_1.2,2.4356268310546874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,4,1,power_law_1.2,0.103189115524292
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,4,1,power_law_1.2,0.1861926460266113
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,4,1,power_law_1.2,0.12391167640686036
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,4,1,power_law_1.2,3.0956332397460935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,4,1,power_law_1.2,0.14365056037902832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,4,1,power_law_1.2,0.40870079040527346
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,4,1,power_law_1.2,0.23057024002075197
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,4,1,power_law_1.2,0.3196665573120117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,4,1,power_law_1.2,0.5820614242553711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,4,1,power_law_1.2,0.7696198272705078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.01,0.07834239959716796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.01,0.09911168098449707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.01,0.14107904434204102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.01,0.28028160095214844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.01,0.38137729644775387
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.01,0.5072422409057618
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.01,0.6820198059082031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.01,0.7120313262939453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.01,0.7252326202392578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.01,0.7308863830566407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.01,0.7657446289062501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,4,1,power_law_1.2,1.1273587036132813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.01,0.8131391906738281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.01,0.8363654327392578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.01,0.8879666900634765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.01,0.9471641540527344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.01,0.9575878143310547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.01,0.9849766540527345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.01,1.082003173828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.01,1.1430815887451171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.01,1.3080601501464844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.01,1.46712890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.01,1.7938233947753908
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.01,2.1426457214355468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.01,2.863367004394531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.01,3.592041625976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.01,5.068615112304688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.01,6.539527587890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.01,0.04645823955535889
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.01,0.07109568119049073
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.01,0.1014521598815918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,4,1,power_law_1.01,8.04498291015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.01,0.1717036819458008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.01,0.2502092742919922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.01,0.3384908676147461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.01,0.4346527862548828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.01,0.46235393524169927
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.01,0.4717728042602539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.01,0.4759603118896484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.01,0.4944390487670899
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.01,0.5071001434326172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.01,0.5141516876220703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.01,0.5234412765502929
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.01,0.5795455932617187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.01,0.5822118377685547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.01,0.6092345428466797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.01,0.6401516723632812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.01,0.6730989074707031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.01,0.7458841705322266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.01,0.8153363037109376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.01,0.9508767700195312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,4,1,power_law_1.2,1.4932934570312502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.01,1.1129548645019531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.01,1.4165895080566406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.01,1.7465426635742187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.01,2.381720275878906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.01,2.996201477050781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,4,1,power_law_1.01,3.6502239990234373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.01,0.059392638206481926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.01,0.07405119895935058
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.01,0.12045503616333009
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.01,0.1604307174682617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.01,0.08870400428771973
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.01,0.19607679367065428
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.01,0.2701708793640137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.01,0.28743295669555663
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,4,1,power_law_1.2,1.8357817077636718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.01,0.29351680755615234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.01,0.2958559989929199
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.01,0.314778881072998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.01,0.3380524826049805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.01,0.33063999176025394
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.01,0.3570304107666016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.01,0.3613228988647461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.01,0.38606273651123046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.01,0.32476478576660156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.01,0.4295391845703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.01,0.5678521728515624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.01,0.4745228958129883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.01,0.6549523162841797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.01,0.8473286437988282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.01,1.0667628479003906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.01,1.4886265563964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.01,1.9401338195800784
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,4,1,power_law_1.2,0.07820799827575683
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,4,1,power_law_1.01,2.835701904296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,4,1,power_law_1.2,0.14165696144104004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,4,1,power_law_1.2,0.09910400390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,4,1,power_law_1.2,0.24897087097167966
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,4,1,power_law_1.2,0.36441280364990236
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,4,1,power_law_1.01,3.7107653808593755
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,4,1,power_law_1.2,0.47693504333496095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,4,1,power_law_1.2,0.6679058837890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,4,1,power_law_1.2,0.7173273468017578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,4,1,power_law_1.2,0.7173971557617188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,4,1,power_law_1.2,0.7368716430664063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,4,1,power_law_1.2,0.7645465850830078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,4,1,power_law_1.01,4.557751770019531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,4,1,power_law_1.2,0.8099411010742188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,4,1,power_law_1.2,0.8460755157470704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,4,1,power_law_1.2,0.9217708587646485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,4,1,power_law_1.2,0.9555750274658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,4,1,power_law_1.2,0.9717625427246095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,4,1,power_law_1.2,1.0144915008544921
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,4,1,power_law_1.2,1.1123187255859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,4,1,power_law_1.2,1.1710399627685546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,4,1,power_law_1.2,1.3431021118164064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,4,1,power_law_1.2,1.5059954833984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,4,1,power_law_1.2,1.840791015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,4,1,power_law_1.2,2.196630401611328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,4,1,power_law_1.2,2.9040231323242187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,4,1,power_law_1.2,3.6489157104492187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,4,1,power_law_1.2,0.04959936141967773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,4,1,power_law_1.2,0.06994815826416015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,4,1,power_law_1.2,5.135166015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,4,1,power_law_1.2,0.10190208435058594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,4,1,power_law_1.2,0.1612646484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,4,1,power_law_1.2,0.23698303222656253
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,4,1,power_law_1.2,0.3276512145996094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,4,1,power_law_1.2,0.427946891784668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,4,1,power_law_1.2,0.4463199996948243
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,4,1,power_law_1.2,0.4634009552001953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,4,1,power_law_1.2,6.612489013671874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,4,1,power_law_1.2,0.4781766510009765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,4,1,power_law_1.2,0.48979198455810546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,4,1,power_law_1.2,0.503430404663086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,4,1,power_law_1.2,0.5154323196411132
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,4,1,power_law_1.2,0.5337120056152344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,4,1,power_law_1.2,0.5849734497070312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,4,1,power_law_1.2,0.6020140838623047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,4,1,power_law_1.2,8.0889453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,4,1,power_law_1.2,0.6195481491088868
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,4,1,power_law_1.2,0.6621408081054687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,4,1,power_law_1.2,0.6994451141357422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,4,1,power_law_1.2,0.7757196807861328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,4,1,power_law_1.2,0.8397132873535156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,4,1,power_law_1.2,0.9990777587890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,4,1,power_law_1.2,1.160668182373047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,4,1,power_law_1.2,1.4545330810546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,4,1,power_law_1.2,1.775992889404297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,4,1,power_law_1.2,2.409660186767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,4,1,power_law_1.2,0.059343361854553224
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,4,1,power_law_1.2,3.052496032714844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,4,1,power_law_1.2,0.07384128093719482
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,4,1,power_law_1.2,0.08665920257568359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,4,1,power_law_1.2,0.11416895866394043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,4,1,power_law_1.2,0.149367036819458
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,4,1,power_law_1.2,3.651856689453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,4,1,power_law_1.2,0.18747583389282227
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,4,1,power_law_1.2,0.26777727127075196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,4,1,power_law_1.2,0.2784000015258789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,4,1,power_law_1.2,0.2896224021911621
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,4,1,power_law_1.2,0.30036863327026364
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,4,1,power_law_1.2,0.3122656059265137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,4,1,power_law_1.2,0.3220614242553711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,4,1,power_law_1.2,0.3382489776611328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,4,1,power_law_1.2,0.3531443023681641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,4,1,power_law_1.2,0.36930686950683594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,4,1,power_law_1.2,0.3833715057373047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,4,1,power_law_1.2,0.41672000885009763
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,4,1,power_law_1.2,0.47682239532470705
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,4,1,power_law_1.2,0.5520972824096679
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,4,1,power_law_1.2,0.7019353485107421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,4,1,power_law_1.2,0.8474752044677734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,4,1,power_law_1.2,1.1617011260986327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,4,1,power_law_1.2,1.5106271362304686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,4,1,power_law_1.2,2.2401394653320312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,4,1,power_law_1.2,2.989377136230469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.01,0.0620684814453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.01,0.0575705623626709
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.01,0.07544511795043946
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.01,0.07537087917327881
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.01,0.07701568126678467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.01,0.07923264026641845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.01,0.08021568298339844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.01,0.08524928092956544
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.01,0.0933471965789795
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.01,0.0948960018157959
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.01,0.0977894401550293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.01,0.10117695808410646
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.01,0.10325632095336915
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.01,0.1093984031677246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.01,0.12779775619506836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.01,0.1502239990234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.01,0.17373760223388673
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.01,0.20432640075683595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.01,0.26707712173461917
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.01,0.3308755111694336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.01,0.4557414245605469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.01,0.5829888153076171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.01,0.8380947113037109
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.01,1.0921619415283204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,4,1,power_law_1.01,1.3469734191894531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.01,0.03161664009094238
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.01,0.03291071891784668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.01,0.041448321342468265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.01,0.04780096054077149
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.01,0.056124157905578606
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.01,0.07570303916931152
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.01,0.07806208133697509
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.01,0.07904704093933106
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.01,0.08085824012756347
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.01,0.08146495819091797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.01,0.08223936080932617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.01,0.09257856369018555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.01,0.0945747184753418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.01,0.09580096244812011
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.01,0.09820032119750977
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.01,0.09988479614257813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.01,0.10700032234191895
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.01,0.11141695976257324
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.01,0.1265651226043701
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.01,0.1469222354888916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.01,0.17082880020141603
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.01,0.21514944076538084
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.01,0.26199104309082033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.01,0.3561254501342773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.01,0.44593471527099615
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.01,0.6361369705200195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.01,0.819304962158203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,4,1,power_law_1.01,1.0035372924804686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.01,0.038278400897979736
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.01,0.0490553617477417
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.01,0.05693183898925781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.01,0.05945087909698486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.01,0.059929599761962896
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.01,0.06146944046020508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.01,0.06159615993499756
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.01,0.0634438419342041
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.01,0.06457151889801024
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.01,0.06573056221008301
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.01,0.0680076789855957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.01,0.03901376008987427
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.01,0.044060158729553225
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.01,0.041543679237365724
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.01,0.07097343921661377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.01,0.07123007774353027
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.01,0.07650815963745117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.01,0.08760383605957031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.01,0.10125375747680665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.01,0.12480256080627442
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.01,0.15172096252441408
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.01,0.20709632873535155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.01,0.2646841621398926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.01,0.3835647964477539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.01,0.504516487121582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.01,0.7591513824462891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2048,768,8,128,4,1,power_law_1.2,0.025093119144439697
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2048,768,8,128,4,1,power_law_1.2,0.04040639877319336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2048,768,8,128,4,1,power_law_1.2,0.047224321365356446
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2048,768,8,128,4,1,power_law_1.2,0.054837760925292967
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2048,768,8,128,4,1,power_law_1.2,0.07613632202148438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2048,768,8,128,4,1,power_law_1.2,0.03771967887878418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2048,768,8,128,4,1,power_law_1.2,0.0753427219390869
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2048,768,8,128,4,1,power_law_1.2,0.07902336120605469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2048,768,8,128,4,1,power_law_1.2,0.0806611156463623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2048,768,8,128,4,1,power_law_1.2,0.0817414379119873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2048,768,8,128,4,1,power_law_1.2,0.08591232299804688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.01,1.0211968231201172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2048,768,8,128,4,1,power_law_1.2,0.09482879638671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2048,768,8,128,4,1,power_law_1.2,0.09619263648986817
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2048,768,8,128,4,1,power_law_1.2,0.09944767951965332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2048,768,8,128,4,1,power_law_1.2,0.10269696235656738
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2048,768,8,128,4,1,power_law_1.2,0.10457344055175781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2048,768,8,128,4,1,power_law_1.2,0.11075519561767577
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2048,768,8,128,4,1,power_law_1.2,0.1339084815979004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2048,768,8,128,4,1,power_law_1.2,0.1756096076965332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2048,768,8,128,4,1,power_law_1.2,0.14971967697143557
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2048,768,8,128,4,1,power_law_1.2,0.20459327697753907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2048,768,8,128,4,1,power_law_1.2,0.26794368743896485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2048,768,8,128,4,1,power_law_1.2,0.3313894271850586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,4,1,power_law_1.01,1.2811814880371093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2048,768,8,128,4,1,power_law_1.2,0.45798591613769535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2048,768,8,128,4,1,power_law_1.2,0.5843148803710937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,2048,768,8,128,4,1,power_law_1.2,0.03126015901565552
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,2048,768,8,128,4,1,power_law_1.2,0.03326911926269531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,2048,768,8,128,4,1,power_law_1.2,0.04074751853942871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,2048,768,8,128,4,1,power_law_1.2,0.046833920478820804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,2048,768,8,128,4,1,power_law_1.2,0.05512447834014893
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,2048,768,8,128,4,1,power_law_1.2,0.07684415817260742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,2048,768,8,128,4,1,power_law_1.2,0.07819583892822266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,2048,768,8,128,4,1,power_law_1.2,0.08076095581054688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,2048,768,8,128,4,1,power_law_1.2,0.08200832366943359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,2048,768,8,128,4,1,power_law_1.2,0.08190912246704102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,2048,768,8,128,4,1,power_law_1.2,0.08236736297607422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,2048,768,8,128,4,1,power_law_1.2,0.09345727920532226
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,2048,768,8,128,4,1,power_law_1.2,0.09518912315368652
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,2048,768,8,128,4,1,power_law_1.2,0.09752063751220703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,2048,768,8,128,4,1,power_law_1.2,0.0999788761138916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,2048,768,8,128,4,1,power_law_1.2,0.10165184020996094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,2048,768,8,128,4,1,power_law_1.2,0.10753984451293945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,2048,768,8,128,4,1,power_law_1.2,0.11439423561096192
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,2048,768,8,128,4,1,power_law_1.2,0.1308512020111084
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,2048,768,8,128,4,1,power_law_1.2,0.15162624359130858
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,2048,768,8,128,4,1,power_law_1.2,0.1741164779663086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,2048,768,8,128,4,1,power_law_1.2,0.21638656616210938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,2048,768,8,128,4,1,power_law_1.2,0.26539775848388675
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2048,768,8,128,4,1,power_law_1.2,0.8413362884521485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,2048,768,8,128,4,1,power_law_1.2,0.35766273498535156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,2048,768,8,128,4,1,power_law_1.2,0.44657535552978517
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2048,768,8,128,4,1,power_law_1.2,1.0964173126220704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,2048,768,8,128,4,1,power_law_1.2,0.6378028869628907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,2048,768,8,128,4,1,power_law_1.2,0.03858752012252807
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,2048,768,8,128,4,1,power_law_1.2,0.040259199142456056
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,2048,768,8,128,4,1,power_law_1.2,0.04294271945953369
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,2048,768,8,128,4,1,power_law_1.2,0.04444863796234131
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,2048,768,8,128,4,1,power_law_1.2,0.04814784049987793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,2048,768,8,128,4,1,power_law_1.2,0.05719744205474854
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,2048,768,8,128,4,1,power_law_1.2,0.05948287963867187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,2048,768,8,128,4,1,power_law_1.2,0.06093183994293213
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,2048,768,8,128,4,1,power_law_1.2,0.06193920135498047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,2048,768,8,128,4,1,power_law_1.2,0.062270078659057614
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,2048,768,8,128,4,1,power_law_1.2,0.0639251184463501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,2048,768,8,128,4,1,power_law_1.2,0.06613311767578126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,2048,768,8,128,4,1,power_law_1.2,0.06684544086456298
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,2048,768,8,128,4,1,power_law_1.2,0.07067776203155518
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,2048,768,8,128,4,1,power_law_1.2,0.07441279888153077
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,2048,768,8,128,4,1,power_law_1.2,0.0775603199005127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2048,768,8,128,4,1,power_law_1.2,1.3534559631347656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,2048,768,8,128,4,1,power_law_1.2,0.08523584365844726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,2048,768,8,128,4,1,power_law_1.2,0.8206854248046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,2048,768,8,128,4,1,power_law_1.2,0.10237055778503419
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,2048,768,8,128,4,1,power_law_1.2,0.12167296409606934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,2048,768,8,128,4,1,power_law_1.2,0.16273088455200196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,2048,768,8,128,4,1,power_law_1.2,0.20464704513549803
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,2048,768,8,128,4,1,power_law_1.2,0.2928166389465332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,2048,768,8,128,4,1,power_law_1.2,0.38494335174560546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,2048,768,8,128,4,1,power_law_1.2,1.0070182037353514
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,2048,768,8,128,4,1,power_law_1.2,0.5668684768676757
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,2048,768,8,128,4,1,power_law_1.2,0.7363283538818359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.01,0.03588608026504517
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.01,0.06256959915161134
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.01,0.08060864448547364
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.01,0.10737343788146973
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.01,0.13184127807617188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.01,0.20895360946655273
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.01,0.21313663482666018
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.01,0.21744640350341796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.01,0.22190784454345702
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.01,0.222926082611084
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.01,0.23355392456054686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.01,0.25755775451660157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.01,0.260053768157959
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.01,0.2659321594238281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.01,0.2701574325561523
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.01,0.27514944076538084
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.01,0.298085765838623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.01,0.3428646469116211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.01,0.3674291229248047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.01,0.4472671890258789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.01,0.5262348937988282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.01,0.6957215881347656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.01,0.8653382110595704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.01,1.2149664306640626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,2048,768,8,128,4,1,power_law_1.2,1.0948326110839843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.01,1.56400634765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.01,2.2621522521972657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.01,2.9595904541015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.01,0.036196479797363276
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.01,0.04801983833312988
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.01,0.06545536041259765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.01,0.08645759582519531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.01,0.11171199798583983
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.01,0.1513862419128418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,4,1,power_law_1.01,3.6590771484374995
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.01,0.15385536193847654
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.01,0.15664064407348632
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.01,0.16044288635253906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.01,0.16102975845336914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.01,0.1646758460998535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.01,0.1856608009338379
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.01,0.18685375213623048
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.01,0.1921900749206543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.01,0.19359872817993165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.01,0.19809024810791015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.01,0.20662912368774414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.01,0.22136831283569336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.01,0.24418880462646486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.01,0.2783481597900391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.01,0.33043712615966797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.01,0.4124313735961914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.01,0.49126846313476563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.01,0.6715302276611328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.01,0.8389676666259765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.01,1.1886131286621093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.01,1.5335813903808595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,4,1,power_law_1.01,1.8815461730957033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,2048,768,8,128,4,1,power_law_1.2,1.4589747619628906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.01,0.037813758850097655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.01,0.042725119590759274
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.01,0.06138688087463379
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.01,0.0710214376449585
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.01,0.09461183547973633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.01,0.0994048023223877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.01,0.10134207725524902
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.01,0.10396415710449218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.01,0.10487104415893556
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.01,0.10829824447631835
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.01,0.11113471984863281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.01,0.11359999656677247
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.01,0.12003456115722658
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.01,0.05168191909790039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.01,0.1272499179840088
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.01,0.13112383842468261
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.01,0.14499327659606934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,2048,768,8,128,4,1,power_law_1.2,1.8289286804199218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.01,0.16835519790649414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.01,0.1941267204284668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.01,0.254836483001709
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.01,0.3145017623901367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.01,0.4390233612060547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.01,0.5651814270019531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.01,0.8421036529541016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.01,1.1056800079345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,4,1,power_law_1.2,0.04515967845916748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,4,1,power_law_1.2,0.06261375904083252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,4,1,power_law_1.2,0.07927680015563965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,4,1,power_law_1.2,0.1009100818634033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,4,1,power_law_1.2,0.12440128326416015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,4,1,power_law_1.2,0.21148351669311521
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.01,1.6427320861816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,4,1,power_law_1.2,0.21387136459350584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,4,1,power_law_1.2,0.22214464187622068
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,4,1,power_law_1.2,0.22524351119995117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,4,1,power_law_1.2,0.22355840682983397
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,4,1,power_law_1.2,0.23572479248046876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.01,2.218756408691406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,4,1,power_law_1.2,0.25741376876831057
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,4,1,power_law_1.2,0.26446207046508785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,4,1,power_law_1.2,0.26621887207031253
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,4,1,power_law_1.2,0.2700435256958008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,4,1,power_law_1.2,0.27666879653930665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,4,1,power_law_1.2,0.2946937561035156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,4,1,power_law_1.2,0.344582405090332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,4,1,power_law_1.2,0.37186752319335936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.01,2.8778298950195316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,4,1,power_law_1.2,0.4598092651367187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,4,1,power_law_1.2,0.5281887817382812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,4,1,power_law_1.2,0.7010438537597656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,4,1,power_law_1.2,0.87384765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,4,1,power_law_1.2,1.222919692993164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,4,1,power_law_1.2,1.573654327392578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,4,1,power_law_1.2,0.034990079402923584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,4,1,power_law_1.2,0.0474732780456543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,4,1,power_law_1.2,0.06330880165100097
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,4,1,power_law_1.2,0.07715199947357178
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,4,1,power_law_1.2,0.10638848304748535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,4,1,power_law_1.2,0.14627455711364745
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,4,1,power_law_1.2,0.15502400398254396
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,4,1,power_law_1.2,0.15745471954345702
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,4,1,power_law_1.2,0.1605926322937012
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,4,1,power_law_1.2,0.16144575119018553
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,4,1,power_law_1.2,0.16500799179077147
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,4,1,power_law_1.2,0.18569984436035156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,4,1,power_law_1.2,0.1871833610534668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,4,1,power_law_1.2,0.19337600708007813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,4,1,power_law_1.2,0.1970163154602051
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,4,1,power_law_1.2,0.19859264373779298
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,4,1,power_law_1.2,0.20810752868652344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,4,1,power_law_1.2,0.22374912261962893
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,4,1,power_law_1.2,0.2501427268981934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,4,1,power_law_1.2,0.2870976066589356
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,4,1,power_law_1.2,0.33258239746093754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,4,1,power_law_1.2,0.4126496124267578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,4,1,power_law_1.2,0.500120964050293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,4,1,power_law_1.2,2.2725599670410155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,4,1,power_law_1.2,0.675251235961914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,4,1,power_law_1.2,0.8463942718505859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,4,1,power_law_1.2,2.9674041748046873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,4,1,power_law_1.2,1.1917990112304688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,4,1,power_law_1.2,0.037801599502563475
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,4,1,power_law_1.2,0.04294847965240479
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,4,1,power_law_1.2,0.05108223915100098
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,4,1,power_law_1.2,0.05883200168609619
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,4,1,power_law_1.2,0.06824512004852294
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,4,1,power_law_1.2,0.0951296043395996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,4,1,power_law_1.2,0.10011136054992675
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,4,1,power_law_1.2,0.10323840141296388
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,4,1,power_law_1.2,0.10550720214843749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,4,1,power_law_1.2,0.10670080184936523
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,4,1,power_law_1.2,0.11001983642578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,4,1,power_law_1.2,0.11428671836853029
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,4,1,power_law_1.2,0.11726719856262206
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,4,1,power_law_1.2,0.12620927810668944
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,4,1,power_law_1.2,0.1343289566040039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,4,1,power_law_1.2,0.1415513610839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,4,1,power_law_1.2,0.16005376815795896
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,4,1,power_law_1.2,3.6630035400390626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,4,1,power_law_1.2,0.1956883239746094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,4,1,power_law_1.2,1.5452038574218752
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,4,1,power_law_1.2,0.23455871582031249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,4,1,power_law_1.2,0.3231916809082031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,4,1,power_law_1.2,0.41063617706298833
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,4,1,power_law_1.2,0.5990316772460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,4,1,power_law_1.2,1.8868666076660154
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,4,1,power_law_1.2,0.7976313781738281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,4,1,power_law_1.2,1.1821094512939454
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,4,1,power_law_1.2,1.55972412109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.01,0.07948736190795899
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.01,0.10073087692260743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.01,0.15145471572875976
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.01,0.2577440071105957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.01,0.33003456115722657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.01,0.5009011077880859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.01,0.5175827026367188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.01,0.5436531066894531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.01,0.5678406524658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.01,0.5774496078491211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.01,0.6019884872436523
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.01,0.6356179046630859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.01,0.6656320190429688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.01,0.72268798828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.01,0.7346195220947266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.01,0.7366355133056641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.01,0.7672223663330078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.01,0.835638427734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.01,0.9125702667236328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.01,1.0660838317871093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.01,1.2550540924072267
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.01,1.622437744140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,4,1,power_law_1.2,2.3290240478515623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.01,1.9956134033203123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.01,2.7738714599609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.01,3.5594860839843747
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.01,5.111313171386719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.01,6.679953002929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.01,0.05340352058410645
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.01,0.07393280029296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.01,0.11821184158325196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.01,0.15707776069641113
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.01,0.21618560791015623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.01,0.32276096343994143
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.01,0.32948097229003903
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.01,0.33954753875732424
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,4,1,power_law_1.01,8.245909423828126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.01,0.34819774627685546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.01,0.35156993865966796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.01,0.3599699020385742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.01,0.3698809432983398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.01,0.409240951538086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.01,0.4255564880371094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.01,0.42694335937499994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.01,0.4173510360717773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.01,0.4490694427490235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.01,0.4744140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.01,0.5060691070556641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.01,0.5679929733276368
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.01,0.6570304107666016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.01,0.8052787017822265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.01,0.9556665802001953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.01,1.2763270568847656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.01,1.5812454223632812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.01,2.186135711669922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,4,1,power_law_1.2,3.0945031738281252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.01,2.8064913940429688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,4,1,power_law_1.01,3.4230316162109373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.01,0.04585343837738037
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.01,0.10080256462097167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.01,0.06345151901245118
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.01,0.12629759788513184
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.01,0.18983104705810547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.01,0.20033983230590818
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.01,0.07700799942016602
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.01,0.2087398338317871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.01,0.21482496261596679
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.01,0.21789440155029297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.01,0.224649600982666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.01,0.23194047927856448
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.01,0.2346944046020508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.01,0.2474105644226074
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.01,0.254465274810791
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.01,0.25690303802490233
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,4,1,power_law_1.2,3.873365783691406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.01,0.28022592544555663
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.01,0.3199654388427734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.01,0.3576057434082031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.01,0.43941310882568363
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.01,0.5337286376953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.01,0.7260665893554688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.01,0.9200601959228516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.01,1.3364518737792968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.01,1.7482508850097656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.01,2.607818908691406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,4,1,power_law_1.2,0.09964096069335937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,4,1,power_law_1.2,0.14743488311767577
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,4,1,power_law_1.2,0.22773248672485352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,4,1,power_law_1.2,0.31524415969848635
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.01,3.498234252929687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,4,1,power_law_1.2,0.48814014434814457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,4,1,power_law_1.2,0.5117030334472656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,4,1,power_law_1.2,0.5464985656738282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,4,1,power_law_1.2,0.5795820617675782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,4,1,power_law_1.2,0.6003705596923827
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,4,1,power_law_1.2,0.6231353759765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,4,1,power_law_1.2,0.07871744155883789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,4,1,power_law_1.2,0.6565939331054688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,4,1,power_law_1.2,0.6909292602539063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,4,1,power_law_1.2,0.7332390594482422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,4,1,power_law_1.2,0.7436275482177734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,4,1,power_law_1.2,0.7579980468749999
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,4,1,power_law_1.2,0.7818386840820313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,4,1,power_law_1.2,0.8606438446044923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,4,1,power_law_1.2,0.920293731689453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,4,1,power_law_1.2,1.0931686401367187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,4,1,power_law_1.2,1.2713555145263673
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,4,1,power_law_1.2,1.6550355529785157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,4,1,power_law_1.2,2.025490570068359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,4,1,power_law_1.2,2.8063058471679687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,4,1,power_law_1.2,3.5855935668945316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,4,1,power_law_1.01,4.41623291015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,4,1,power_law_1.2,0.05848256111145019
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,4,1,power_law_1.2,0.07295040130615235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,4,1,power_law_1.2,0.11143168449401855
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,4,1,power_law_1.2,0.14624832153320313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,4,1,power_law_1.2,0.20742591857910156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,4,1,power_law_1.2,0.31793472290039065
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,4,1,power_law_1.2,0.32747329711914064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,4,1,power_law_1.2,0.33984001159667965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,4,1,power_law_1.2,0.3525196838378906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,4,1,power_law_1.2,0.3579372787475586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,4,1,power_law_1.2,5.146487426757813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,4,1,power_law_1.2,0.3686393737792969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,4,1,power_law_1.2,0.3727283096313477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,4,1,power_law_1.2,0.4254374313354492
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,4,1,power_law_1.2,0.4285919952392578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,4,1,power_law_1.2,0.4394028854370117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,4,1,power_law_1.2,0.44651264190673834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,4,1,power_law_1.2,0.46156223297119137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,4,1,power_law_1.2,0.4799731063842773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,4,1,power_law_1.2,0.5208147048950196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,4,1,power_law_1.2,0.5917075347900391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,4,1,power_law_1.2,0.6686003112792969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,4,1,power_law_1.2,0.8237593841552734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,4,1,power_law_1.2,0.9757401275634766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,4,1,power_law_1.2,1.2869541931152344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,4,1,power_law_1.2,1.5966514587402343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,4,1,power_law_1.2,6.7113037109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,4,1,power_law_1.2,2.2066163635253906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,4,1,power_law_1.2,0.04600768089294434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,4,1,power_law_1.2,0.06025919914245605
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,4,1,power_law_1.2,0.07508031845092773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,4,1,power_law_1.2,0.09412991523742675
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,4,1,power_law_1.2,0.11985471725463867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,4,1,power_law_1.2,0.18724800109863282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,4,1,power_law_1.2,0.19857023239135743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,4,1,power_law_1.2,0.20959936141967775
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,4,1,power_law_1.2,0.21865663528442383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,4,1,power_law_1.2,0.22334720611572267
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,4,1,power_law_1.2,2.8480133056640624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,4,1,power_law_1.2,0.23181631088256838
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,4,1,power_law_1.2,8.269814453124999
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,4,1,power_law_1.2,0.23646272659301756
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,4,1,power_law_1.2,0.24702720642089843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,4,1,power_law_1.2,0.25581247329711915
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,4,1,power_law_1.2,3.443034362792969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,4,1,power_law_1.2,0.28325887680053713
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,4,1,power_law_1.2,0.27141056060791013
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,4,1,power_law_1.2,0.3082271957397461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,4,1,power_law_1.2,0.3626079940795898
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,4,1,power_law_1.2,0.42276672363281254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,4,1,power_law_1.2,0.5581356811523437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,4,1,power_law_1.2,0.7042483520507813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,4,1,power_law_1.2,1.0219654083251952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,4,1,power_law_1.2,1.3648281860351563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,4,1,power_law_1.2,2.0716966247558597
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,4,1,power_law_1.2,2.7881787109375002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.01,0.08161984443664551
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.01,0.1028927993774414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.01,0.14574591636657713
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.01,0.29740352630615235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.01,0.46770175933837893
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.01,0.6316275024414063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.01,0.7421241760253906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.01,0.9303673553466798
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.01,0.9627085113525391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.01,0.9961555480957032
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.01,1.0338623809814453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.01,1.0517330932617188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.01,1.0968646240234374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.01,1.1570816040039062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,4,1,power_law_1.2,4.260507507324219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.01,1.2058425903320313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.01,1.314812774658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.01,1.290847930908203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.01,1.4180633544921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.01,1.4836441040039063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.01,1.654163818359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.01,1.8089549255371093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.01,2.155132141113281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.01,2.5116883850097658
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.01,3.20593017578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.01,3.9059243774414063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.01,5.386371459960937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.01,0.05315072059631347
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.01,0.07477439880371094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.01,0.10609215736389159
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.01,0.17650560379028318
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.01,0.30029375076293946
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.01,0.4205785751342773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,4,1,power_law_1.01,8.388751220703124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.01,0.48860736846923825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.01,0.5945606231689453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.01,0.619892463684082
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.01,0.6467858886718749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,4,1,power_law_1.2,5.7519946289062505
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.01,0.6645005035400391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.01,0.6743353271484376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.01,0.6953913879394531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.01,0.7158592224121094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.01,0.7242438507080078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.01,0.7496428680419922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.01,0.8179122924804687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.01,0.8513823699951171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.01,0.8876140594482422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.01,0.9664390563964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.01,1.0436646270751955
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.01,1.192208023071289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.01,1.3490655517578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.01,1.678948516845703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.01,1.9928160095214842
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.01,2.6522125244140624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.01,3.289969177246094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,4,1,power_law_1.01,3.9730438232421874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.01,0.07383296012878418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.01,0.0764415979385376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.01,0.07427264213562011
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.01,0.07524352073669434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.01,0.08078720092773437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.01,0.08697279930114746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.01,0.0991744041442871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.01,0.10184896469116211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.01,0.11184191703796387
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.01,0.12459391593933104
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.01,0.1382361602783203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.01,0.18390655517578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.01,0.2470457649230957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.01,0.20672191619873045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.01,0.29821760177612305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.01,0.3408563232421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.01,0.42647743225097656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.01,0.6058329772949219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.01,1.154596481323242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.01,0.7800045013427734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.01,1.5102426147460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.01,2.2428671264648434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.01,2.9462322998046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,4,1,power_law_1.2,0.08980095863342286
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,4,1,power_law_1.2,0.10297856330871583
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.01,4.39122314453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,4,1,power_law_1.2,0.14409472465515136
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,4,1,power_law_1.2,0.4025075149536133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,4,1,power_law_1.2,0.5754291152954101
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,4,1,power_law_1.2,0.6727308654785157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,4,1,power_law_1.2,0.9094226837158204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,4,1,power_law_1.2,0.9380051422119141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,4,1,power_law_1.2,0.9599929809570312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,4,1,power_law_1.2,0.2843635177612305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,4,1,power_law_1.2,0.9953683471679687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,4,1,power_law_1.2,1.0524966430664062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,4,1,power_law_1.2,1.087583999633789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,4,1,power_law_1.2,1.1557689666748048
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,4,1,power_law_1.2,1.2257465362548827
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,4,1,power_law_1.2,1.2996095275878905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,4,1,power_law_1.2,1.338439636230469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,4,1,power_law_1.2,1.5028218078613282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,4,1,power_law_1.2,1.6837811279296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,4,1,power_law_1.2,1.4350982666015626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,4,1,power_law_1.2,1.8663545227050782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,4,1,power_law_1.2,2.227161560058594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,4,1,power_law_1.2,2.5769607543945314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,4,1,power_law_1.2,3.273555908203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,4,1,power_law_1.2,4.00084228515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,4,1,power_law_1.2,5.488973388671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,4,1,power_law_1.2,7.00267578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,4,1,power_law_1.2,0.05236991882324219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,4,1,power_law_1.2,0.07396160125732422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,4,1,power_law_1.2,0.1045689582824707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,4,1,power_law_1.2,0.17409280776977537
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,4,1,power_law_1.2,0.264836483001709
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,4,1,power_law_1.2,8.501519165039062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,4,1,power_law_1.2,0.3807187271118164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,4,1,power_law_1.2,0.45346561431884763
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,4,1,power_law_1.2,0.585412483215332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,4,1,power_law_1.2,0.6059436798095703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,4,1,power_law_1.2,0.6231129455566407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,4,1,power_law_1.2,0.6468179321289063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,4,1,power_law_1.2,0.6697811126708985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,4,1,power_law_1.2,0.6880255889892578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,4,1,power_law_1.2,0.7091187286376953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,4,1,power_law_1.2,0.7243981170654297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,4,1,power_law_1.2,0.7437420654296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,4,1,power_law_1.2,0.822007064819336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,4,1,power_law_1.2,0.8688928222656249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,4,1,power_law_1.2,0.9150182342529296
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,4,1,power_law_1.2,0.9989683532714844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,4,1,power_law_1.2,1.086903076171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,4,1,power_law_1.2,1.2486700439453124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,4,1,power_law_1.2,1.4097491455078124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,4,1,power_law_1.2,1.7314784240722656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,4,1,power_law_1.2,2.0584063720703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,4,1,power_law_1.2,2.7082437133789066
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,4,1,power_law_1.2,0.07419456005096435
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,4,1,power_law_1.2,3.3594650268554687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,4,1,power_law_1.2,0.07539328098297118
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,4,1,power_law_1.2,0.0836678409576416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,4,1,power_law_1.2,0.09107135772705079
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,4,1,power_law_1.2,4.008137512207031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,4,1,power_law_1.2,0.0977734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,4,1,power_law_1.2,0.11471743583679199
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,4,1,power_law_1.2,0.12496512413024903
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,4,1,power_law_1.2,0.1457484817504883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,4,1,power_law_1.2,0.18452224731445313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,4,1,power_law_1.2,0.20525056838989258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,4,1,power_law_1.2,0.2474483108520508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,4,1,power_law_1.2,0.29931264877319336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,4,1,power_law_1.2,0.07869952201843262
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,4,1,power_law_1.2,0.34085056304931644
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,4,1,power_law_1.2,0.07346496105194092
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,4,1,power_law_1.2,0.10262144088745118
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,4,1,power_law_1.2,0.4276972961425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,4,1,power_law_1.2,0.6026444625854492
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,4,1,power_law_1.2,0.7819878387451171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,4,1,power_law_1.2,1.1523372650146484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,4,1,power_law_1.2,1.5143104553222657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,4,1,power_law_1.2,2.231756134033203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,4,1,power_law_1.2,2.9628839111328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,4,1,power_law_1.2,4.382035827636718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.01,0.0596288013458252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.01,0.08140928268432616
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.01,0.10213760375976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.01,0.16235008239746093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.01,0.16642047882080077
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.01,0.16964288711547854
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.01,0.17344768524169923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.01,0.1749945640563965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.01,0.18385215759277346
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.01,0.19564352035522461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.01,0.19956287384033203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.01,0.2041152000427246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.01,0.20600191116333008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.01,0.2118560028076172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.01,0.23873600006103515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.01,0.2540838432312012
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.01,0.2727187156677246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.01,0.32379966735839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.01,0.3699065780639649
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.01,0.47484161376953127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,4,1,power_law_1.01,0.586627197265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.01,0.8091020965576172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.01,1.032126693725586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.01,1.4779385375976564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.01,1.9262322998046877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,4,1,power_law_1.01,2.3738143920898436
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,4,1,power_law_1.2,0.05868095874786376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,4,1,power_law_1.2,0.07549056053161621
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,4,1,power_law_1.2,0.09573311805725097
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,4,1,power_law_1.2,0.16076288223266602
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,4,1,power_law_1.2,0.16601472854614258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,4,1,power_law_1.2,0.17038015365600584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,4,1,power_law_1.2,0.172108154296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,4,1,power_law_1.2,0.17614656448364258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,4,1,power_law_1.2,0.02842303991317749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,4,1,power_law_1.2,0.1907161521911621
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,4,1,power_law_1.2,0.2017919921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,4,1,power_law_1.2,0.2068179130554199
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,4,1,power_law_1.2,0.21003776550292969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,4,1,power_law_1.2,0.19596351623535155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,4,1,power_law_1.2,0.2308742332458496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,4,1,power_law_1.2,0.0421017599105835
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,4,1,power_law_1.2,0.23945791244506837
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,4,1,power_law_1.2,0.2567526435852051
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,4,1,power_law_1.2,0.33184959411621096
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,4,1,power_law_1.2,0.37652671813964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,4,1,power_law_1.2,0.48414337158203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,4,1,power_law_1.2,0.8157984161376953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,4,1,power_law_1.2,0.2711091232299805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,4,1,power_law_1.2,1.0430834960937498
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,4,1,power_law_1.2,1.4890713500976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,4,1,power_law_1.2,1.9321510314941406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,4,1,power_law_1.2,2.386504364013672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.01,0.09476608276367188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.01,0.2750707244873047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.01,0.1681964874267578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.01,1.011447067260742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.01,0.45043582916259767
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.01,0.6274617767333984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.01,1.0843218994140624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.01,1.1688953399658204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.01,1.1909932708740234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.01,1.1265555572509764
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.01,1.245401611328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.01,1.3140646362304689
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.01,1.3857664489746093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.01,1.577487335205078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.01,1.726929931640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.01,1.9219187927246093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.01,2.289105224609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.01,1.6503033447265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.01,1.791658935546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.01,2.1079437255859377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.01,2.4427033996582033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.01,3.135859069824219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.01,3.8517739868164065
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.01,5.359378662109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.01,6.886909790039063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.01,0.0783135986328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.01,9.923377075195313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.01,0.11486783981323243
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.01,0.28422208786010744
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.01,0.39990142822265623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.01,0.18580032348632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.01,0.6241676712036133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.01,12.95109619140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.01,0.6952326202392578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.01,0.7218675231933593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.01,0.7303673553466796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.01,0.7505152130126953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.01,0.7738438415527343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.01,0.7852729797363281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.01,0.8180921936035157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.01,0.8139743804931641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.01,0.6702201843261719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,4,1,power_law_1.01,16.014837646484374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.01,0.8444416046142578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.01,0.8879257965087891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.01,1.1136511993408202
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.01,1.2655014038085937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.01,1.4211987304687501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.01,1.0447923278808593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.01,1.7541766357421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.01,2.0716511535644533
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.01,2.724684143066406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.01,3.388778991699219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.01,4.756251525878906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.01,6.135032958984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,4,1,power_law_1.01,7.463436889648437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,power_law_1.01,0.1087724781036377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,power_law_1.01,0.11315711975097656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,power_law_1.01,0.10870207786560057
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,power_law_1.01,0.10944704055786132
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,power_law_1.01,0.12076736450195313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,power_law_1.01,0.11036479949951172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,power_law_1.01,0.12704511642456057
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,power_law_1.01,0.1351206398010254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,power_law_1.01,0.14158656120300292
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,power_law_1.01,0.14867775917053222
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,power_law_1.01,0.16139392852783202
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,power_law_1.01,0.19429376602172851
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,power_law_1.01,0.2091084861755371
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,power_law_1.01,0.2368876838684082
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,power_law_1.01,0.2818239974975586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,power_law_1.01,0.31111423492431645
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,power_law_1.01,0.38716991424560543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,power_law_1.01,0.6794182586669921
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,power_law_1.01,0.5314598464965821
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,power_law_1.01,0.9766022491455079
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,power_law_1.01,1.270238723754883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,power_law_1.01,1.852986297607422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,power_law_1.01,2.439430999755859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,power_law_1.01,3.6304473876953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,4,1,power_law_1.2,0.10495295524597167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,power_law_1.01,4.794439697265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,4,1,power_law_1.2,0.16897855758666994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,4,1,power_law_1.2,0.27777599334716796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,4,1,power_law_1.2,0.418737907409668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,4,1,power_law_1.2,0.5620748901367187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,4,1,power_law_1.2,0.9940518188476564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,4,1,power_law_1.2,1.081229476928711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,4,1,power_law_1.2,1.115959701538086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,4,1,power_law_1.2,1.164007034301758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,4,1,power_law_1.2,1.2115577697753905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,4,1,power_law_1.2,1.2733535766601562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,4,1,power_law_1.2,1.3612991333007813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,4,1,power_law_1.2,1.4435084533691405
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,4,1,power_law_1.2,1.6377888488769532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,4,1,power_law_1.2,1.8080351257324216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,4,1,power_law_1.2,1.9885215759277344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,4,1,power_law_1.2,2.3538514709472658
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,4,1,power_law_1.2,1.7049530029296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,4,1,power_law_1.2,1.8508460998535157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,4,1,power_law_1.2,2.1547251892089845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,4,1,power_law_1.2,2.489890594482422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,4,1,power_law_1.2,3.1971392822265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,4,1,power_law_1.2,3.9297714233398438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,4,1,power_law_1.2,5.43861083984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,4,1,power_law_1.2,6.9666650390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,4,1,power_law_1.2,9.993611450195313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,4,1,power_law_1.2,0.0744595193862915
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,4,1,power_law_1.2,0.18147455215454103
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,4,1,power_law_1.2,0.26460735321044926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,4,1,power_law_1.2,0.37113471984863283
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,4,1,power_law_1.2,0.6206252670288086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,4,1,power_law_1.2,0.6687999725341797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,4,1,power_law_1.2,0.6870816040039063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,4,1,power_law_1.2,0.11018879890441895
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,4,1,power_law_1.2,0.7183686065673828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,4,1,power_law_1.2,0.761440658569336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,4,1,power_law_1.2,13.049404296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,4,1,power_law_1.2,0.7843090820312499
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,4,1,power_law_1.2,0.7346112060546874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,4,1,power_law_1.2,0.8036953735351563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,4,1,power_law_1.2,0.8457350158691407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,4,1,power_law_1.2,0.8446822357177733
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,4,1,power_law_1.2,0.8747392272949218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,4,1,power_law_1.2,0.9222662353515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,4,1,power_law_1.2,1.158657913208008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,4,1,power_law_1.2,1.3090675354003907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,4,1,power_law_1.2,1.4684255981445313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,4,1,power_law_1.2,1.0909037017822265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,4,1,power_law_1.2,1.7903648376464845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,4,1,power_law_1.2,2.113414459228516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,4,1,power_law_1.2,2.7812127685546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,4,1,power_law_1.2,3.4468255615234376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,4,1,power_law_1.2,4.8198239135742185
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,2688,22,512,4,1,power_law_1.2,0.1093113613128662
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,2688,22,512,4,1,power_law_1.2,0.10896575927734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,2688,22,512,4,1,power_law_1.2,0.10913536071777344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,2688,22,512,4,1,power_law_1.2,0.11012096405029297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,2688,22,512,4,1,power_law_1.2,0.11299839973449707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,4,1,power_law_1.2,16.080135498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,2688,22,512,4,1,power_law_1.2,0.12041152000427247
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,2688,22,512,4,1,power_law_1.2,0.12732799530029298
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,4,1,power_law_1.2,6.1568670654296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,2688,22,512,4,1,power_law_1.2,0.1345414352416992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,2688,22,512,4,1,power_law_1.2,0.14168255805969238
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,2688,22,512,4,1,power_law_1.2,0.16280319213867187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,2688,22,512,4,1,power_law_1.2,0.19411455154418947
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,2688,22,512,4,1,power_law_1.2,0.2097420883178711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,2688,22,512,4,1,power_law_1.2,0.23467327117919923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,2688,22,512,4,1,power_law_1.2,0.2835929679870605
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,2688,22,512,4,1,power_law_1.2,0.14833279609680178
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,4,1,power_law_1.2,7.517100219726562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,2688,22,512,4,1,power_law_1.2,0.31171455383300783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,2688,22,512,4,1,power_law_1.2,0.388139533996582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,2688,22,512,4,1,power_law_1.2,0.5317715072631837
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,2688,22,512,4,1,power_law_1.2,0.6786605072021484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,2688,22,512,4,1,power_law_1.2,0.9723795318603516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,2688,22,512,4,1,power_law_1.2,1.2733875274658204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,2688,22,512,4,1,power_law_1.2,2.4384921264648436
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,2688,22,512,4,1,power_law_1.2,1.8524333190917968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,2688,22,512,4,1,power_law_1.2,3.6181140136718755
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,2688,22,512,4,1,power_law_1.2,4.814241333007812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.01,0.07165823936462402
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.01,0.06661888122558593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.01,0.06997695922851563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.01,0.08656703948974609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.01,0.11401023864746093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.01,0.1155724811553955
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.01,0.11678079605102538
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.01,0.11686528205871582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.01,0.1174687957763672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.01,0.11946368217468262
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.01,0.12348031997680664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.01,0.12487232208251953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.01,0.12955904006958008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.01,0.13481280326843262
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.01,0.13888832092285158
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.01,0.1685753631591797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.01,0.2011020851135254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.01,0.23159423828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.01,0.3542617416381836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.01,0.43516990661621097
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.01,0.6062995147705078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.01,0.7772140502929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.01,1.122375717163086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.01,1.467349090576172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.01,2.1571615600585936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.01,0.036505599021911625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.01,0.042028160095214845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.01,0.0448089599609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.01,0.059032320976257324
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.01,0.06229631900787354
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.01,0.06313471794128418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.01,0.06390207767486572
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.01,0.06373248100280762
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.01,0.06407487869262696
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.01,0.06528063774108886
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.01,0.06676352024078369
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.01,0.06644288063049317
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.01,0.06911359786987305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.01,0.0756774377822876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.01,0.08191679954528809
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.01,0.09012479782104492
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.01,0.09152000427246094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.01,0.09738752365112305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.01,0.11644927978515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.01,0.17310400009155275
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.01,0.2292927932739258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.01,0.13973183631896974
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,8,1,power_law_1.01,3.5361587524414064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.01,0.2989638328552246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.01,0.41498046874999994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.01,0.5444313430786133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.01,0.7795116424560546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.01,1.0090585327148438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.01,0.018943359851837156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.01,0.034076159000396726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.01,0.035329279899597166
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.01,0.035660159587860105
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.01,0.03570175886154175
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.01,0.034906880855560304
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.01,0.034958720207214355
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.01,0.0360806393623352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.01,0.036170239448547366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,8,1,power_law_1.01,1.2577945709228515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.01,0.021386239528656006
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.01,0.024904320240020754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.01,0.03652096033096314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.01,0.03769279956817627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.01,0.03758975982666016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.01,0.04087679862976074
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.01,0.04376575946807861
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.01,0.04724160194396972
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.01,0.053360638618469236
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.01,0.06287871837615967
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.01,0.07944384098052978
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.01,0.09488832473754882
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.01,0.12869312286376952
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.01,0.16186368942260743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.01,0.2297599983215332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.01,0.2997433662414551
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,8,1,power_law_1.2,0.0565664005279541
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,8,1,power_law_1.2,0.06528831958770752
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,8,1,power_law_1.2,0.06818431854248047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,8,1,power_law_1.2,0.0868832015991211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,8,1,power_law_1.2,0.11201727867126465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,8,1,power_law_1.2,0.11492032051086425
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,8,1,power_law_1.2,0.1164691162109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,8,1,power_law_1.2,0.11663423538208009
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,8,1,power_law_1.2,0.11788800239562988
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,8,1,power_law_1.2,0.11904512405395509
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,8,1,power_law_1.2,0.12372608184814453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,8,1,power_law_1.2,0.12526016235351561
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,8,1,power_law_1.2,0.12905983924865722
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,8,1,power_law_1.2,0.1350111961364746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,8,1,power_law_1.2,0.13934720039367676
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,8,1,power_law_1.2,0.1806380844116211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,8,1,power_law_1.2,0.20146432876586914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,8,1,power_law_1.2,0.23113855361938476
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,8,1,power_law_1.2,0.2764588737487793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.01,0.4345529556274414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,8,1,power_law_1.2,0.35765441894531247
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,8,1,power_law_1.2,0.4351628875732422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,8,1,power_law_1.2,0.6073088073730469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.01,0.5695840072631836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,8,1,power_law_1.2,0.773114242553711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,8,1,power_law_1.2,1.1194822692871094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.01,0.7050656127929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,8,1,power_law_1.2,1.461432342529297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,8,1,power_law_1.2,0.03926208019256592
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,8,1,power_law_1.2,0.04527808189392089
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,8,1,power_law_1.2,0.04681727886199951
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,8,1,power_law_1.2,0.05862912178039551
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,8,1,power_law_1.2,0.06100671768188477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,8,1,power_law_1.2,0.062060160636901854
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,8,1,power_law_1.2,0.06376959800720214
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,8,1,power_law_1.2,0.06318336009979249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,8,1,power_law_1.2,0.06498112201690673
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,8,1,power_law_1.2,0.0643500804901123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,8,1,power_law_1.2,0.06630208015441894
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,8,1,power_law_1.2,0.06682432174682618
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,8,1,power_law_1.2,0.06799104213714599
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,8,1,power_law_1.2,0.07473087787628174
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,8,1,power_law_1.2,0.08336576461791992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,8,1,power_law_1.2,0.09415871620178222
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,8,1,power_law_1.2,0.09151167869567871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,8,1,power_law_1.2,0.097455997467041
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,8,1,power_law_1.2,0.11681344032287597
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,8,1,power_law_1.2,0.14030271530151367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,8,1,power_law_1.2,0.1732703971862793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,8,1,power_law_1.2,2.1501324462890627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,8,1,power_law_1.2,0.2320217514038086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,8,1,power_law_1.2,0.29815488815307617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,8,1,power_law_1.2,0.4205868911743164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,8,1,power_law_1.2,2.8422860717773437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,8,1,power_law_1.2,0.5390227127075196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,8,1,power_law_1.2,3.524369812011719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,8,1,power_law_1.2,0.019728000164031985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,8,1,power_law_1.2,0.022609920501708985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,8,1,power_law_1.2,0.02382911920547485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,8,1,power_law_1.2,0.03493695974349976
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,8,1,power_law_1.2,0.034572160243988036
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,8,1,power_law_1.2,0.036417920589447025
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,8,1,power_law_1.2,0.03637056112289429
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,8,1,power_law_1.2,0.03598975896835327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,8,1,power_law_1.2,0.03667327880859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,8,1,power_law_1.2,0.03691263914108277
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,8,1,power_law_1.2,0.03709248065948486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,8,1,power_law_1.2,0.03719935894012451
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,8,1,power_law_1.2,0.036974079608917236
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,8,1,power_law_1.2,0.03907776117324829
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,8,1,power_law_1.2,0.041462402343749996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,8,1,power_law_1.2,0.04513279914855957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,8,1,power_law_1.2,0.7771526336669922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,8,1,power_law_1.2,0.04712512016296387
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,8,1,power_law_1.2,0.05568895816802979
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,8,1,power_law_1.2,1.0197401428222657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,8,1,power_law_1.2,0.0973356819152832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,8,1,power_law_1.2,0.06293248176574708
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,8,1,power_law_1.2,0.07976704120635987
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,8,1,power_law_1.2,1.2482598114013672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,8,1,power_law_1.2,0.13384960174560548
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,8,1,power_law_1.2,0.16636735916137696
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,8,1,power_law_1.2,0.23855360031127928
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,8,1,power_law_1.2,0.3073273658752441
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.01,0.07392191886901855
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.01,0.08783424377441407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.01,0.09640576362609862
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.01,0.1273420810699463
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.01,0.16240831375122072
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.01,0.16320831298828126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.01,0.16374847412109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.01,0.16391807556152344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.01,0.1642323112487793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.01,0.1662668800354004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.01,0.17276287078857422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.01,0.18395135879516603
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.01,0.19500736236572264
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.01,0.20758079528808593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.01,0.22367935180664061
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.01,0.265152645111084
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.01,0.2937830352783203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.01,0.3578470230102539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.01,0.42547073364257815
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.01,0.5641363143920899
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.01,0.704136962890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.01,0.9891699218750001
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.01,1.2763308715820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,8,1,power_law_1.2,0.44940032958984377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.01,1.8527052307128904
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.01,2.4294169616699217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,8,1,power_law_1.2,0.5867552185058594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.01,3.578853759765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.01,0.04919360160827636
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.01,0.0584012794494629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.01,0.061767678260803226
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.01,0.08113151550292969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.01,0.0857926368713379
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.01,0.08760383605957031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.01,0.08862079620361328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.01,0.08866687774658202
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.01,0.08902272224426269
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.01,0.09040448188781738
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.01,0.09187135696411133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.01,0.09294976234436035
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.01,0.09466752052307129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.01,0.10789376258850099
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.01,0.1130297565460205
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.01,0.11499391555786134
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.01,0.12282048225402833
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.01,0.13616191864013671
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.01,0.16931392669677733
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.01,0.2117305564880371
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.01,0.24342720031738282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.01,4.738780212402344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.01,0.3400044631958008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.01,0.4340780639648438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,8,1,power_law_1.2,0.7262348937988281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.01,0.6043609619140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.01,0.7888172912597656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,8,1,power_law_1.01,6.052205810546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.01,1.1415955352783205
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.01,1.4898054504394531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.01,0.02444159984588623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.01,0.04688511848449707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.01,0.04724991798400879
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.01,0.047698559761047366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.01,0.04761216163635254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.01,0.04756159782409668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.01,0.047204480171203614
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.01,0.04814591884613037
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.01,0.04879424095153809
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,8,1,power_law_1.01,1.8627980041503907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.01,0.030430719852447507
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.01,0.03787328004837036
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.01,0.05006400108337402
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.01,0.05139455795288086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.01,0.0554694414138794
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.01,0.05776959896087647
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.01,0.058360958099365236
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.01,0.06387648105621338
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.01,0.07630464076995849
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.01,0.08871040344238282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.01,0.11286975860595702
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.01,0.13957951545715333
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.01,0.18878143310546874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.01,0.23864959716796874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.01,0.33974143981933597
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.01,0.4455168151855469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,8,1,power_law_1.2,0.07445248126983642
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,8,1,power_law_1.2,0.08758784294128417
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,8,1,power_law_1.2,0.09212032318115235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,8,1,power_law_1.2,0.12518272399902344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,8,1,power_law_1.2,0.1553536033630371
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,8,1,power_law_1.2,0.1593337631225586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,8,1,power_law_1.2,0.16013504028320313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,8,1,power_law_1.2,0.16044416427612304
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,8,1,power_law_1.2,0.1620332717895508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,8,1,power_law_1.2,0.1642822456359863
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,8,1,power_law_1.2,0.17601600646972657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,8,1,power_law_1.2,0.18582336425781248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,8,1,power_law_1.2,0.1954854393005371
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,8,1,power_law_1.2,0.20927104949951172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,8,1,power_law_1.2,0.21880064010620118
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,8,1,power_law_1.2,0.27865856170654296
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,8,1,power_law_1.2,0.29389568328857424
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.01,0.6483615875244141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,8,1,power_law_1.2,0.3584646224975586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,8,1,power_law_1.2,0.4311756896972656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,8,1,power_law_1.2,0.5682278442382812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,8,1,power_law_1.2,0.7040019226074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,8,1,power_law_1.2,0.9927353668212892
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,8,1,power_law_1.2,1.2742720031738282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.01,0.8491622161865233
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,8,1,power_law_1.2,1.850518341064453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.01,1.0628582763671877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,8,1,power_law_1.2,2.420868530273437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,8,1,power_law_1.2,0.04922304153442383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,8,1,power_law_1.2,0.056306557655334474
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,8,1,power_law_1.2,0.05918528079986572
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,8,1,power_law_1.2,0.08191807746887207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,8,1,power_law_1.2,0.08515199661254882
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,8,1,power_law_1.2,0.08766079902648925
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,8,1,power_law_1.2,0.08860032081604004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,8,1,power_law_1.2,0.088919038772583
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,8,1,power_law_1.2,0.09043904304504394
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,8,1,power_law_1.2,0.0907686424255371
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,8,1,power_law_1.2,0.09230848312377929
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,8,1,power_law_1.2,0.09335359573364257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,8,1,power_law_1.2,0.09416128158569335
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,8,1,power_law_1.2,0.10741760253906249
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,8,1,power_law_1.2,0.11700351715087891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,8,1,power_law_1.2,0.11839103698730469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,8,1,power_law_1.2,0.1251353645324707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,8,1,power_law_1.2,0.13606143951416017
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,8,1,power_law_1.2,0.16915584564208985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,8,1,power_law_1.2,0.2024736022949219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,8,1,power_law_1.2,0.24496959686279296
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,8,1,power_law_1.2,3.5737152099609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,8,1,power_law_1.2,0.34036544799804686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,8,1,power_law_1.2,0.42818878173828123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,8,1,power_law_1.2,0.6072140884399414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,8,1,power_law_1.2,4.734016723632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,8,1,power_law_1.2,0.7806265258789062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,8,1,power_law_1.2,6.2588220214843755
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,8,1,power_law_1.2,0.024418559074401856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,8,1,power_law_1.2,0.0293503999710083
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,8,1,power_law_1.2,0.032821760177612305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,8,1,power_law_1.2,0.04673408031463623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,8,1,power_law_1.2,0.046671361923217775
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,8,1,power_law_1.2,0.04746431827545166
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,8,1,power_law_1.2,0.04788671970367432
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,8,1,power_law_1.2,0.0472537612915039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,8,1,power_law_1.2,0.04811135768890381
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,8,1,power_law_1.2,0.04815936088562012
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,8,1,power_law_1.2,0.048763518333435056
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,8,1,power_law_1.2,0.05043263912200928
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,8,1,power_law_1.2,0.05071231842041015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,8,1,power_law_1.2,0.05556159973144531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,8,1,power_law_1.2,0.059260158538818365
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,8,1,power_law_1.2,0.060407681465148924
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,8,1,power_law_1.2,1.1394726562500002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,8,1,power_law_1.2,0.07715072154998778
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,8,1,power_law_1.2,0.06441472053527833
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,8,1,power_law_1.2,0.08996416091918945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,8,1,power_law_1.2,1.5078668212890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,8,1,power_law_1.2,0.11410431861877442
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,8,1,power_law_1.2,1.8412498474121093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,8,1,power_law_1.2,0.1407583999633789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,8,1,power_law_1.2,0.1951263999938965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,8,1,power_law_1.2,0.24396736145019532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,8,1,power_law_1.2,0.350748176574707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,8,1,power_law_1.2,0.4558003234863281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.01,0.0649561595916748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.01,0.08070079803466797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.01,0.10204416275024415
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.01,0.1453587245941162
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.01,0.21242176055908205
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.01,0.30622272491455077
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.01,0.3906291198730469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.01,0.3993094253540039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.01,0.4036742401123047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.01,0.4082336044311523
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.01,0.42716289520263667
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.01,0.45891456604003905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.01,0.48158016204833987
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.01,0.5067731094360352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.01,0.5588505554199219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.01,0.5915372848510743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.01,0.6295865631103515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.01,0.6607641601562501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.01,0.6985804748535156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.01,0.7928300476074218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,8,1,power_law_1.2,0.6698201751708984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.01,0.9051129913330078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.01,1.1034099578857421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.01,1.327772216796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.01,1.782027587890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.01,2.2305555725097657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.01,3.1395904541015627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.01,4.048035888671874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.01,0.04481984138488769
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.01,0.05398848056793213
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.01,0.07987840175628662
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.01,0.12044416427612306
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.01,0.18301183700561524
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.01,0.24224639892578126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,8,1,power_law_1.01,4.974642028808594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.01,0.30915327072143556
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.01,0.3200454330444336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.01,0.32489662170410155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.01,0.33079742431640624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,8,1,power_law_1.2,0.8789753723144532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.01,0.341473274230957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.01,0.35224254608154293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.01,0.35714561462402344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.01,0.36452545166015626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.01,0.4097958374023437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.01,0.41393409729003905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.01,0.435681266784668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.01,0.46550014495849606
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.01,0.49258239746093746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.01,0.5434950256347657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.01,0.6107110214233399
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.01,0.7192499542236328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.01,0.8470579528808594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.01,1.1025228881835938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.01,1.3690483093261718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.01,1.8702297973632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.01,2.376728973388672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,8,1,power_law_1.01,2.892736511230469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.01,0.060314879417419434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.01,0.1005945587158203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.01,0.0660313606262207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.01,0.07998720169067383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.01,0.1268070411682129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.01,0.1498431968688965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.01,0.18959936141967773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.01,0.19942335128784178
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.01,0.20440256118774416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.01,0.20779392242431644
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.01,0.21665279388427736
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.01,0.22378751754760745
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.01,0.2273151969909668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.01,0.23531839370727542
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.01,0.2464588737487793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.01,0.2541567993164063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.01,0.272289924621582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,8,1,power_law_1.2,1.0861363220214844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.01,0.3122879981994629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.01,0.34748096466064454
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.01,0.43801345825195315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.01,0.5252588653564454
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.01,0.6969273376464844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.01,0.8830355072021485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.01,1.2778182220458985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.01,1.7073587036132813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,8,1,power_law_1.2,0.06373248100280762
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,8,1,power_law_1.2,0.08025152206420898
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,8,1,power_law_1.2,0.10189824104309082
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,8,1,power_law_1.2,0.13872768402099608
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,8,1,power_law_1.2,0.19633472442626954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,8,1,power_law_1.2,0.28780992507934566
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,8,1,power_law_1.2,0.38282688140869137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,8,1,power_law_1.2,0.39415233612060546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,8,1,power_law_1.2,0.4006476974487304
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,8,1,power_law_1.2,0.4098694229125977
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,8,1,power_law_1.2,0.4255667114257813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,8,1,power_law_1.2,0.4613056182861328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,8,1,power_law_1.2,0.4842476654052734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,8,1,power_law_1.2,0.5215513610839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,8,1,power_law_1.2,0.5715020751953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,8,1,power_law_1.2,0.6141196823120116
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,8,1,power_law_1.2,0.6336249542236329
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,8,1,power_law_1.2,0.672972183227539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,8,1,power_law_1.01,2.574871826171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,8,1,power_law_1.01,3.3482791137695314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,8,1,power_law_1.2,0.7078534698486327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,8,1,power_law_1.2,0.8256409454345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,8,1,power_law_1.2,0.917093734741211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,8,1,power_law_1.2,1.1332268524169922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,8,1,power_law_1.01,4.093766479492187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,8,1,power_law_1.2,1.3513848876953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,8,1,power_law_1.2,1.7972166442871091
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,8,1,power_law_1.2,2.250019226074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,8,1,power_law_1.2,0.046418561935424804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,8,1,power_law_1.2,3.1608920288085938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,8,1,power_law_1.2,0.054344959259033196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,8,1,power_law_1.2,0.08006848335266113
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,8,1,power_law_1.2,0.1143712043762207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,8,1,power_law_1.2,4.080781555175781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,8,1,power_law_1.2,0.1758355140686035
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,8,1,power_law_1.2,0.23289152145385744
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,8,1,power_law_1.2,0.30639871597290036
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,8,1,power_law_1.2,0.31586624145507813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,8,1,power_law_1.2,0.32206207275390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,8,1,power_law_1.2,0.3305215835571289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,8,1,power_law_1.2,0.3388915252685547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,8,1,power_law_1.2,0.3495782470703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,8,1,power_law_1.2,0.35854976654052734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,8,1,power_law_1.2,0.37202239990234376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,8,1,power_law_1.2,0.41406974792480467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,8,1,power_law_1.2,0.4295462417602539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,8,1,power_law_1.2,0.444692497253418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,8,1,power_law_1.2,4.994194641113281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,8,1,power_law_1.2,0.47935745239257815
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,8,1,power_law_1.2,0.5058201599121094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,8,1,power_law_1.2,0.5675807952880859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,8,1,power_law_1.2,0.6290764617919922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,8,1,power_law_1.2,0.7604601287841797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,8,1,power_law_1.2,0.8890284729003906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,8,1,power_law_1.2,1.1183283233642578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,8,1,power_law_1.2,1.3865945434570313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,8,1,power_law_1.2,1.8873779296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,8,1,power_law_1.2,0.060220799446105956
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,8,1,power_law_1.2,2.414241943359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,8,1,power_law_1.2,0.07788159847259521
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,8,1,power_law_1.2,0.09743231773376465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,8,1,power_law_1.2,2.9106585693359373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,8,1,power_law_1.2,0.12013504028320313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,8,1,power_law_1.2,0.14437055587768555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,8,1,power_law_1.2,0.1873708724975586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,8,1,power_law_1.2,0.19578943252563477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,8,1,power_law_1.2,0.20231487274169924
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,8,1,power_law_1.2,0.20898752212524413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,8,1,power_law_1.2,0.21674303054809568
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,8,1,power_law_1.2,0.22468671798706055
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,8,1,power_law_1.2,0.2320351982116699
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,8,1,power_law_1.2,0.24525375366210939
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,8,1,power_law_1.2,0.25883136749267577
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,8,1,power_law_1.2,0.27062847137451174
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,8,1,power_law_1.2,0.30064256668090816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,8,1,power_law_1.2,0.41677696228027344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,8,1,power_law_1.2,0.06685056209564208
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,8,1,power_law_1.2,0.3592639923095703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,8,1,power_law_1.2,0.5598508834838867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,8,1,power_law_1.2,0.6990137481689453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,8,1,power_law_1.2,0.9920992279052735
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,8,1,power_law_1.2,1.338260498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,8,1,power_law_1.2,1.989122619628906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,8,1,power_law_1.2,2.7214974975585937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.01,0.051799039840698245
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.01,0.03449023962020874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.01,0.07351871967315673
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.01,0.11576319694519044
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.01,0.12078847885131835
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.01,0.1245689582824707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.01,0.12961728096008301
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.01,0.13141247749328613
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.01,0.14373311996459961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.01,0.15682944297790527
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.01,0.15976063728332518
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.01,0.08747200012207032
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.01,0.166296329498291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.01,0.16824703216552733
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.01,0.17316671371459963
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.01,0.18276800155639647
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.01,0.2145145606994629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.01,0.2611321640014649
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.01,0.3039699172973633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.01,0.3603276824951172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.01,0.4748972702026367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.01,0.5898303985595703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.01,0.8183923339843749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.01,1.048892822265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.01,1.5110176086425782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.01,1.975811767578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,8,1,power_law_1.01,2.439374694824219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.01,0.034891519546508785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.01,0.04231040000915527
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.01,0.06733503818511963
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.01,0.0860051155090332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.01,0.1155942440032959
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.01,0.12217535972595214
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.01,0.12483455657958983
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.01,0.05490240097045899
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.01,0.1265459156036377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.01,0.12818047523498535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.01,0.13107839584350586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.01,0.14749631881713868
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.01,0.1496230411529541
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.01,0.1538368034362793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.01,0.15804672241210938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.01,0.1588755226135254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.01,0.16864896774291993
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.01,0.18157888412475587
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.01,0.2062214469909668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.01,0.23936639785766603
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.01,0.27902975082397463
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.01,0.3545356750488281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.01,0.4306284713745117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.01,0.5899430465698242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.01,0.7374674987792968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.01,1.0533939361572267
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.01,1.35935302734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,8,1,power_law_1.01,1.6709222412109377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.01,0.04105984210968018
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.01,0.04887104034423828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.01,0.04399807929992676
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.01,0.06419775962829591
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.01,0.08034624099731445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.01,0.08446463584899902
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.01,0.08591679573059083
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.01,0.08803775787353516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.01,0.08909952163696289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.01,0.09175104141235352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.01,0.09428992271423339
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.01,0.09698687553405762
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.01,0.10251968383789063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.01,0.05530687808990479
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.01,0.10798015594482421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.01,0.11204480171203612
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.01,0.1209670352935791
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.01,0.14349887847900392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.01,0.17123008728027345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.01,0.22487871170043947
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.01,0.2742950439453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.01,0.3837401580810547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.01,0.49609153747558593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.01,0.7527487945556641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.01,0.9829459381103515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.01,1.4735813903808594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,8,1,power_law_1.2,0.03370431900024414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.01,1.9999545288085936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,8,1,power_law_1.2,0.0714028787612915
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,8,1,power_law_1.2,0.08363264083862305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,8,1,power_law_1.2,0.11786239624023438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.01,2.5753689575195313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,8,1,power_law_1.2,0.12255231857299805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,8,1,power_law_1.2,0.12734463691711426
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,8,1,power_law_1.2,0.13227904319763184
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,8,1,power_law_1.2,0.13338175773620606
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,8,1,power_law_1.2,0.14427776336669923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,8,1,power_law_1.2,0.15736576080322268
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,8,1,power_law_1.2,0.16228160858154297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,8,1,power_law_1.2,0.06253183841705322
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,8,1,power_law_1.2,0.16582464218139648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,8,1,power_law_1.2,0.1716223907470703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,8,1,power_law_1.2,0.17401983261108397
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,8,1,power_law_1.2,0.18471359252929687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,8,1,power_law_1.2,0.2171340751647949
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,8,1,power_law_1.2,0.26415103912353516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,8,1,power_law_1.2,0.30519168853759765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,8,1,power_law_1.2,0.36111614227294925
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,8,1,power_law_1.2,0.05240064144134522
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,8,1,power_law_1.2,0.47625087738037114
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,8,1,power_law_1.2,0.5929324722290039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,8,1,power_law_1.2,0.8207283020019531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,8,1,power_law_1.2,1.051626205444336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,8,1,power_law_1.2,0.03577791929244995
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,8,1,power_law_1.2,0.04645440101623535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,8,1,power_law_1.2,0.05432384014129639
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,8,1,power_law_1.2,0.06448768138885498
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,8,1,power_law_1.2,0.08405759811401367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,8,1,power_law_1.2,0.11639295578002931
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,8,1,power_law_1.2,0.12371583938598633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,8,1,power_law_1.2,0.12584256172180175
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,8,1,power_law_1.2,0.12864959716796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,8,1,power_law_1.2,0.1294092845916748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,8,1,power_law_1.2,0.13143808364868165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,8,1,power_law_1.2,0.1509324836730957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,8,1,power_law_1.2,0.1516371154785156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,8,1,power_law_1.2,0.15705792427062987
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,8,1,power_law_1.2,0.15957375526428225
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,8,1,power_law_1.2,0.16278528213500976
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,8,1,power_law_1.2,0.16986303329467772
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,8,1,power_law_1.2,0.18647552490234376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,8,1,power_law_1.2,0.2127654457092285
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,8,1,power_law_1.2,1.5176287841796874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,8,1,power_law_1.2,0.24638847351074217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,8,1,power_law_1.2,0.28438976287841794
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,8,1,power_law_1.2,0.35765441894531247
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,8,1,power_law_1.2,0.43573440551757814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,8,1,power_law_1.2,0.5957625579833985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,8,1,power_law_1.2,0.7407353973388672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,8,1,power_law_1.2,1.9801708984375002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,8,1,power_law_1.2,1.0607513427734374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,8,1,power_law_1.2,2.443255615234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,8,1,power_law_1.2,0.04495296001434326
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,8,1,power_law_1.2,0.04852928161621094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,8,1,power_law_1.2,0.054275197982788084
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,8,1,power_law_1.2,0.062300801277160645
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,8,1,power_law_1.2,0.04104191780090332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,8,1,power_law_1.2,0.08067135810852051
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,8,1,power_law_1.2,0.08515904426574708
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,8,1,power_law_1.2,0.08720831871032715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,8,1,power_law_1.2,0.09048576354980468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,8,1,power_law_1.2,0.08935808181762696
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,8,1,power_law_1.2,0.0936620807647705
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,8,1,power_law_1.2,0.09725119590759276
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,8,1,power_law_1.2,0.09997504234313966
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,8,1,power_law_1.2,0.10684608459472655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,8,1,power_law_1.2,0.11442239761352539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,8,1,power_law_1.2,0.1209772777557373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,8,1,power_law_1.2,1.360262451171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,8,1,power_law_1.2,0.16954048156738283
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,8,1,power_law_1.2,0.13748096466064452
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,8,1,power_law_1.2,0.20785600662231446
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,8,1,power_law_1.2,0.287139835357666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,8,1,power_law_1.2,0.3693708801269531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,8,1,power_law_1.2,1.6741976928710938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,8,1,power_law_1.2,0.5425766372680665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,8,1,power_law_1.2,0.7183411407470703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,8,1,power_law_1.2,1.0614189147949218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,8,1,power_law_1.2,1.406622772216797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.01,0.06378496170043946
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.01,0.07916992187500001
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.01,0.10119232177734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.01,0.14007552146911623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.01,0.18586687088012693
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.01,0.32038078308105467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.01,0.3223763275146484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.01,0.32910015106201174
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.01,0.3320121765136719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.01,0.3351148986816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.01,0.34224510192871094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.01,0.360552978515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.01,0.3915584182739258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.01,0.4037919998168945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.01,0.41560127258300783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.01,0.4245439910888672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.01,0.4841439819335937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.01,0.5192588806152344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.01,0.5509376144409179
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.01,0.6371699142456054
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.01,0.7583699035644531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.01,0.9795724487304687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.01,1.2041964721679688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.01,1.6709919738769532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.01,2.138056335449219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.01,3.0678497314453126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,8,1,power_law_1.2,2.1087974548339843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.01,4.004920959472656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.01,0.04557248115539551
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.01,0.06025023937225342
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.01,0.0822316837310791
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,8,1,power_law_1.01,4.935935363769532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.01,0.11051712036132813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.01,0.2290656089782715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.01,0.23063680648803714
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.01,0.23593088150024416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.01,0.16046464920043946
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.01,0.24154687881469727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.01,0.24444799423217772
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.01,0.24850944519042967
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.01,0.25501184463500975
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.01,0.2927801513671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.01,0.30885568618774417
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.01,0.30899200439453123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.01,0.3009286308288574
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.01,0.33197822570800783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.01,0.3573062515258789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.01,0.37718078613281253
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.01,0.44174335479736326
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.01,0.5025888061523437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.01,0.635272331237793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.01,0.7616710662841797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.01,1.0166015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.01,1.2810348510742187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.01,1.7888870239257812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.01,2.3101344299316406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,8,1,power_law_1.01,2.826256103515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,8,1,power_law_1.2,2.8026605224609376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.01,0.04515967845916748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.01,0.08017279624938964
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.01,0.05225855827331542
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.01,0.09728320121765137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.01,0.13613887786865236
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.01,0.14318655967712401
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.01,0.14705216407775878
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.01,0.15083840370178223
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.01,0.15289152145385743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.01,0.158090238571167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.01,0.16385919570922852
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.01,0.16659391403198243
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.01,0.17709375381469727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.01,0.18344959259033203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.01,0.18648832321166992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.01,0.06596608161926269
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,8,1,power_law_1.2,3.498590087890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.01,0.2077484893798828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.01,0.24122175216674804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.01,0.27581695556640623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.01,0.34941825866699217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.01,0.4326700973510742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.01,0.597729263305664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.01,0.7701331329345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.01,1.1262438201904297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.01,1.493441925048828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,8,1,power_law_1.2,0.06356800079345704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.01,2.2520338439941403
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,8,1,power_law_1.2,0.07960127830505372
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,8,1,power_law_1.2,0.1297216033935547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,8,1,power_law_1.2,0.1010643196105957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,8,1,power_law_1.2,0.17355199813842775
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,8,1,power_law_1.2,0.3063680076599121
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,8,1,power_law_1.2,0.32005054473876954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,8,1,power_law_1.2,0.32624767303466795
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,8,1,power_law_1.2,0.33286209106445314
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,8,1,power_law_1.2,0.33645503997802734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,8,1,power_law_1.2,0.3457664108276367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,8,1,power_law_1.2,0.37538047790527346
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,8,1,power_law_1.2,0.39560321807861326
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,8,1,power_law_1.2,0.40635711669921876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,8,1,power_law_1.2,0.44245185852050783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,8,1,power_law_1.2,0.4511577606201172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,8,1,power_law_1.2,0.4924038314819336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,8,1,power_law_1.2,0.5214374542236329
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,8,1,power_law_1.2,0.5529926300048829
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,8,1,power_law_1.2,0.6520588684082032
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,8,1,power_law_1.2,0.7620601654052734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,8,1,power_law_1.2,0.9922412872314453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,8,1,power_law_1.2,1.2172064208984374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,8,1,power_law_1.2,1.6841587829589844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,8,1,power_law_1.2,2.1482701110839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.01,3.037633972167969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,8,1,power_law_1.2,3.0808990478515623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,8,1,power_law_1.2,0.04557951927185059
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,8,1,power_law_1.2,0.061110401153564455
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,8,1,power_law_1.2,0.08096128463745117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,8,1,power_law_1.2,0.1054041576385498
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,8,1,power_law_1.2,0.15494655609130858
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,8,1,power_law_1.01,3.821919250488281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,8,1,power_law_1.2,0.2255615997314453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,8,1,power_law_1.2,0.22969408035278321
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,8,1,power_law_1.2,0.2380166435241699
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,8,1,power_law_1.2,0.24309759140014647
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,8,1,power_law_1.2,0.2475961685180664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,8,1,power_law_1.2,0.2530054473876953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,8,1,power_law_1.2,0.2581043243408203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,8,1,power_law_1.2,0.31061119079589844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,8,1,power_law_1.2,0.31124927520751955
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,8,1,power_law_1.2,0.3229856109619141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,8,1,power_law_1.2,0.32995838165283203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,8,1,power_law_1.2,0.34271873474121095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,8,1,power_law_1.2,0.36084350585937497
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,8,1,power_law_1.2,0.3850553512573242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,8,1,power_law_1.2,0.45264320373535155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,8,1,power_law_1.2,0.518458251953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,8,1,power_law_1.2,0.6509471893310547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,8,1,power_law_1.2,0.7689830780029296
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,8,1,power_law_1.2,4.018533630371094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,8,1,power_law_1.2,1.027267837524414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,8,1,power_law_1.2,1.286417236328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,8,1,power_law_1.2,1.8061843872070313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,8,1,power_law_1.2,0.045180802345275876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,8,1,power_law_1.2,0.052472958564758296
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,8,1,power_law_1.2,0.06355775833129883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,8,1,power_law_1.2,0.07593791961669923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,8,1,power_law_1.2,0.09363455772399902
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,8,1,power_law_1.2,0.1336083221435547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,8,1,power_law_1.2,0.14228992462158202
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,8,1,power_law_1.2,4.959151306152344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,8,1,power_law_1.2,0.1489510440826416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,8,1,power_law_1.2,0.15346816062927246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,8,1,power_law_1.2,0.15655551910400392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,8,1,power_law_1.2,0.16378559112548827
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,8,1,power_law_1.2,2.317994842529297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,8,1,power_law_1.2,0.1681452751159668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,8,1,power_law_1.2,0.17620927810668946
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,8,1,power_law_1.2,0.18438592910766602
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,8,1,power_law_1.2,0.19864959716796876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,8,1,power_law_1.2,0.23225471496582034
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,8,1,power_law_1.2,2.8445330810546876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,8,1,power_law_1.2,0.20944000244140626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,8,1,power_law_1.2,0.2802387237548828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,8,1,power_law_1.2,0.3366880035400391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,8,1,power_law_1.2,0.45687934875488284
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,8,1,power_law_1.2,0.5888230514526367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,8,1,power_law_1.2,0.8699327850341797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,8,1,power_law_1.2,1.1741734313964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,8,1,power_law_1.2,1.8048371887207033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,8,1,power_law_1.2,2.468138885498047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.01,0.06799744129180908
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.01,0.08451583862304687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.01,0.10598527908325195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.01,0.1519859218597412
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.01,0.2793369674682617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.01,0.38310272216796876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.01,0.4132863998413086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.01,0.5212972640991211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,8,1,power_law_1.2,3.7784402465820315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.01,0.5455852890014649
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.01,0.5777299118041992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.01,0.5777056121826172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.01,0.5898419189453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.01,0.6068742370605469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.01,0.6453804779052734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.01,0.6801651000976563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.01,0.7203424072265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.01,0.7815615844726562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.01,0.8253369903564453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.01,0.9162854766845703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.01,0.9965325164794923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.01,1.084439697265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.01,1.3078746032714843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.01,1.548778839111328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.01,2.0014547729492187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,8,1,power_law_1.2,5.11295166015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.01,2.447538604736328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.01,0.04926527976989746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.01,0.05670271873474121
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.01,0.08441408157348633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.01,0.1264409637451172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.01,0.20846464157104494
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.01,0.30804096221923827
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.01,4.289926452636719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.01,0.3496025466918945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.01,0.4206950378417969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,8,1,power_law_1.01,5.2246069335937495
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.01,0.43583614349365235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.01,0.44833984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.01,0.46154945373535156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.01,0.46881473541259766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.01,0.48383102416992185
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.01,0.5033772659301758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.01,0.5074278259277344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.01,0.524695053100586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.01,0.5876095962524415
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.01,0.6117433547973633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.01,0.6391500854492188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.01,0.6988556671142578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.01,0.7719692993164062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.01,0.8912268829345702
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.01,1.0179948425292968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.01,1.29202880859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.01,1.5463296508789062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.01,2.0830015563964843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.01,2.6131027221679686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,8,1,power_law_1.01,3.1454302978515623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.01,0.07551936149597169
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.01,0.07585408210754394
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.01,0.07711679935455322
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.01,0.08079680442810058
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.01,0.07605055809020997
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.01,0.08687871932983399
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.01,0.09315967559814453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.01,0.10423551559448244
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.01,0.10502079963684081
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.01,0.12855232238769532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.01,0.11190400123596192
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.01,0.17009727478027342
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.01,0.18691392898559572
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.01,0.23055744171142578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.01,0.27830015182495116
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.01,0.31982912063598634
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.01,0.40121280670166015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.01,0.5752544021606445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.01,0.7337849426269532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.01,1.0727187347412108
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.01,1.404363555908203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.01,2.1035250854492187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.01,2.7547021484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,8,1,power_law_1.2,0.07074944019317628
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,8,1,power_law_1.2,0.08357631683349609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.01,4.118923645019531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,8,1,power_law_1.2,0.104583683013916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,8,1,power_law_1.2,0.163319034576416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,8,1,power_law_1.2,0.22929407119750977
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,8,1,power_law_1.2,0.3618783950805664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,8,1,power_law_1.2,0.40232318878173834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,8,1,power_law_1.2,0.5112563323974609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,8,1,power_law_1.2,0.5379276657104493
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,8,1,power_law_1.2,0.5547820663452148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,8,1,power_law_1.2,0.5697433471679687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,8,1,power_law_1.2,0.5845676803588867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,8,1,power_law_1.2,0.6043519973754883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,8,1,power_law_1.2,0.6478291320800781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,8,1,power_law_1.2,0.6874905395507812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,8,1,power_law_1.2,0.7255033874511719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,8,1,power_law_1.2,0.781475830078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,8,1,power_law_1.2,0.8648979187011718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,8,1,power_law_1.2,0.9229510498046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,8,1,power_law_1.2,1.007553939819336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,8,1,power_law_1.2,1.1300057220458986
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,8,1,power_law_1.2,1.3421824645996092
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,8,1,power_law_1.2,1.5747161865234376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,8,1,power_law_1.2,2.021222381591797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,8,1,power_law_1.2,2.480083160400391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,8,1,power_law_1.2,3.399848327636719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,8,1,power_law_1.2,4.339827270507813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,8,1,power_law_1.2,0.04804096221923828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,8,1,power_law_1.2,0.054966402053833005
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,8,1,power_law_1.2,0.08311615943908691
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,8,1,power_law_1.2,0.12367232322692871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,8,1,power_law_1.2,5.27535400390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,8,1,power_law_1.2,0.193505916595459
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,8,1,power_law_1.2,0.28454591751098635
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,8,1,power_law_1.2,0.334186897277832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,8,1,power_law_1.2,0.41682369232177735
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,8,1,power_law_1.2,0.42813056945800787
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,8,1,power_law_1.2,0.43781822204589843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,8,1,power_law_1.2,0.4504006576538086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,8,1,power_law_1.2,0.46548225402832033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,8,1,power_law_1.2,0.47868801116943355
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,8,1,power_law_1.2,0.49756607055664065
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,8,1,power_law_1.2,0.5072947311401367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,8,1,power_law_1.2,0.5229817581176758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,8,1,power_law_1.2,0.5857382583618164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,8,1,power_law_1.2,0.6216044616699219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,8,1,power_law_1.2,0.6506201934814453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,8,1,power_law_1.2,0.7288224029541015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,8,1,power_law_1.2,0.7952082824707032
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,8,1,power_law_1.2,0.934405746459961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,8,1,power_law_1.2,1.0699609374999999
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,8,1,power_law_1.2,1.3260928344726561
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,8,1,power_law_1.2,1.5896038818359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,8,1,power_law_1.2,2.1213644409179686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,8,1,power_law_1.2,2.662109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,8,1,power_law_1.2,3.1839065551757812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,8,1,power_law_1.2,0.07589824199676513
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,8,1,power_law_1.2,0.07591296195983886
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,8,1,power_law_1.2,0.07718527793884278
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,8,1,power_law_1.2,0.08085311889648437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,8,1,power_law_1.2,0.09988351821899413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,8,1,power_law_1.2,0.10613887786865235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,8,1,power_law_1.2,0.11352895736694335
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,8,1,power_law_1.2,0.12633472442626953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,8,1,power_law_1.2,0.17314943313598632
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,8,1,power_law_1.2,0.07505663871765136
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,8,1,power_law_1.2,0.19316736221313477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,8,1,power_law_1.2,0.28074176788330074
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,8,1,power_law_1.2,0.3195244789123535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,8,1,power_law_1.2,0.09225728034973144
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,8,1,power_law_1.2,0.08571455955505372
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,8,1,power_law_1.2,0.22998720169067383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,8,1,power_law_1.2,0.4034553527832031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,8,1,power_law_1.2,0.5735007858276366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,8,1,power_law_1.2,0.7323104095458984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,8,1,power_law_1.2,1.0699571228027343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,8,1,power_law_1.2,1.4167564392089844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,8,1,power_law_1.2,2.102033233642578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,8,1,power_law_1.2,2.7570285034179687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,8,1,power_law_1.2,4.102274475097657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.01,0.07367680072784424
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.01,0.08385343551635742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.01,0.10508031845092773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.01,0.15313920021057129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.01,0.15917695999145506
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.01,0.1653945541381836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.01,0.1715577507019043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.01,0.17894079208374022
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.01,0.1980415916442871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.01,0.21140159606933592
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.01,0.21556543350219726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.01,0.22152767181396485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.01,0.2264236831665039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.01,0.2343827247619629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.01,0.2543686485290527
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.01,0.2990796852111816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.01,0.32785343170166015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.01,0.38380481719970705
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.01,0.4554739379882813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.01,0.578202896118164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.01,0.7132883453369141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.01,0.9746406555175782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.01,1.2378182220458984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.01,1.7606605529785155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.01,2.2873004150390623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,8,1,power_law_1.01,2.8105670166015626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,8,1,power_law_1.2,0.03003328084945679
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,8,1,power_law_1.2,0.04723264217376709
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,8,1,power_law_1.2,0.06153535842895508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,8,1,power_law_1.2,0.07819776058197021
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,8,1,power_law_1.2,0.09890048027038574
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,8,1,power_law_1.2,0.1512377643585205
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,8,1,power_law_1.2,0.15936256408691407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,8,1,power_law_1.2,0.16742271423339844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,8,1,power_law_1.2,0.175548152923584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,8,1,power_law_1.2,0.18407424926757812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,8,1,power_law_1.2,0.2013088035583496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,8,1,power_law_1.2,0.21627519607543944
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,8,1,power_law_1.2,0.22052991867065427
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,8,1,power_law_1.2,0.22903936386108398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,8,1,power_law_1.2,0.23496576309204104
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,8,1,power_law_1.2,0.2446976089477539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,8,1,power_law_1.2,0.2672896003723145
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,8,1,power_law_1.2,0.3044147109985352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,8,1,power_law_1.2,0.33117439270019533
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,8,1,power_law_1.2,0.3932070541381836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,8,1,power_law_1.2,0.46046142578125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,8,1,power_law_1.2,0.5854060745239258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,8,1,power_law_1.2,0.7152294158935547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,8,1,power_law_1.2,0.975753631591797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,8,1,power_law_1.2,1.2386310577392579
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,8,1,power_law_1.2,1.764673309326172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.01,0.0770195198059082
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.01,0.1023686408996582
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.01,0.1749849510192871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.01,0.3619942474365234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.01,0.5718745422363282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.01,0.6102188873291016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.01,0.6285798263549804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.01,0.6538285064697266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.01,0.6735801696777344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.01,0.7094745635986328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.01,0.7531622314453126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.01,0.7936434936523438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.01,0.9039180755615235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.01,0.25443904876708984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.01,0.9746790313720703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,8,1,power_law_1.2,2.291340179443359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.01,1.078455047607422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.01,1.2896429443359376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.01,0.9704338836669922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.01,1.0545600128173827
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.01,1.2797254180908202
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.01,1.5040325927734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.01,1.9613562011718748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.01,2.423627471923828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,8,1,power_law_1.2,2.8137158203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.01,3.3901208496093753
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.01,4.347063598632813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.01,0.05830848217010498
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.01,0.0847219181060791
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.01,0.1295967960357666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.01,6.29399658203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.01,0.20242624282836913
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.01,0.2781478309631348
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.01,0.4323321533203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.01,0.4536409759521485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.01,8.204512939453124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.01,0.47142654418945307
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.01,0.49009983062744145
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.01,0.49163326263427737
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.01,0.5079865646362305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.01,0.5260755157470702
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.01,0.5343750381469727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.01,0.5613817596435547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.01,0.5519417572021484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.01,0.5749331283569336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,8,1,power_law_1.01,10.169264526367188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.01,0.6106681442260742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.01,0.7062937927246093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.01,0.7608921813964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.01,0.880615692138672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.01,1.0054035186767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.01,1.2599833679199217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.01,1.5033888244628906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.01,2.0265209960937502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.01,2.531957702636719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.01,3.580621337890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.01,4.6142822265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,8,1,power_law_1.01,5.68271240234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,8,1,power_law_1.2,0.07782847881317138
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,8,1,power_law_1.2,0.17595136642456055
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,8,1,power_law_1.2,0.2512787246704101
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,8,1,power_law_1.2,0.34602367401123046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,8,1,power_law_1.2,0.5717881774902344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,8,1,power_law_1.2,0.6098220825195313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,8,1,power_law_1.2,0.10391231536865235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,8,1,power_law_1.2,0.6228550338745118
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,8,1,power_law_1.2,0.6572921752929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,8,1,power_law_1.2,0.6948953247070312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,8,1,power_law_1.2,0.7223596954345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,8,1,power_law_1.2,0.7766368103027343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,8,1,power_law_1.2,0.8230489349365234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,8,1,power_law_1.2,0.9328262329101562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,8,1,power_law_1.2,1.0177030181884765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,8,1,power_law_1.2,1.1168089294433594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,8,1,power_law_1.2,0.9953209686279297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,8,1,power_law_1.2,1.3242515563964843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,8,1,power_law_1.2,1.0965126037597657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,8,1,power_law_1.2,1.306189422607422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,8,1,power_law_1.2,1.5272377014160157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,8,1,power_law_1.2,1.9916281127929687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,8,1,power_law_1.2,2.45476806640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,8,1,power_law_1.2,3.4255636596679686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,8,1,power_law_1.2,4.385281372070312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,8,1,power_law_1.2,6.3337255859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,8,1,power_law_1.2,8.243705444335937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,8,1,power_law_1.2,0.08414912223815918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,8,1,power_law_1.2,0.12883392333984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,8,1,power_law_1.2,0.18657472610473633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,8,1,power_law_1.2,0.25931711196899415
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,8,1,power_law_1.2,0.42694271087646485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,8,1,power_law_1.2,0.45276416778564454
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,8,1,power_law_1.2,0.4664217758178711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,8,1,power_law_1.2,0.053927040100097655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,8,1,power_law_1.2,0.4886700820922852
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,8,1,power_law_1.2,0.49672191619873046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,8,1,power_law_1.2,0.5148767852783204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,8,1,power_law_1.2,0.5295251083374024
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,8,1,power_law_1.2,0.5442752075195313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,8,1,power_law_1.2,0.5799033737182617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,8,1,power_law_1.2,0.575335693359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,8,1,power_law_1.2,0.5960383987426757
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,8,1,power_law_1.2,0.6325625610351563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,8,1,power_law_1.2,0.732965087890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,8,1,power_law_1.2,0.7871692657470704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,8,1,power_law_1.2,0.9084690856933595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,8,1,power_law_1.2,1.029388122558594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,8,1,power_law_1.2,1.2810342407226563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,8,1,power_law_1.2,2.0507200622558592
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,8,1,power_law_1.2,1.52652099609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,8,1,power_law_1.2,2.5640289306640627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,8,1,power_law_1.2,3.6051205444335936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,8,1,power_law_1.2,10.203427124023438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,8,1,power_law_1.2,4.643430480957031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,8,1,power_law_1.2,5.686439208984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.01,0.046012158393859866
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.01,0.05177408218383789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.01,0.0537824010848999
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.01,0.06131519794464111
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.01,0.0891756820678711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.01,0.0899788761138916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.01,0.09162943840026856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.01,0.09197823524475097
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.01,0.09295295715332032
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.01,0.09416704177856446
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.01,0.09664768218994141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.01,0.09785599708557129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.01,0.10072896003723145
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.01,0.10396863937377929
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.01,0.1080281639099121
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.01,0.1153932762145996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.01,0.11891263961791991
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.01,0.1683302307128906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.01,0.21655807495117188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.01,0.2590111923217774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.01,0.3526899337768555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.01,0.4440742492675781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.01,0.6317107009887695
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.01,0.8218566131591796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.01,1.2003199768066406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.01,0.031264638900756835
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.01,0.033320960998535154
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.01,0.03493376016616821
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.01,0.038641920089721685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.01,0.04296127796173096
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.01,0.043452157974243164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.01,0.04561600208282471
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.01,0.045425281524658204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.01,0.045171198844909666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.01,0.046555519104003906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.01,0.046785922050476075
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.01,0.04773824214935303
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.01,0.05013567924499511
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.01,0.05650112152099609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.01,0.05897984027862548
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.01,0.06341055870056153
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.01,0.06352575778961181
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.01,0.07013440132141113
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.01,0.0849728012084961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.01,0.10887295722961425
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,16,1,power_law_1.01,1.9568511962890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.01,0.12030976295471192
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.01,0.1646758460998535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.01,0.21204671859741211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.01,0.2898105621337891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.01,0.38050750732421873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.01,0.5456665420532227
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.01,0.018139519691467286
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.01,0.019773440361022947
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.01,0.02054271936416626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.01,0.02059072017669678
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.01,0.0209606409072876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.01,0.020972158908843994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.01,0.021022720336914064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.01,0.02100735902786255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.01,0.020969600677490235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.01,0.021368958950042725
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.01,0.7068364715576172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.01,0.018583040237426757
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,16,1,power_law_1.01,0.8680063629150391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.01,0.022617599964141845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.01,0.02574143886566162
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.01,0.023011200428009033
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.01,0.02670655965805054
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.01,0.031587839126586914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.01,0.03942271947860718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.01,0.03475136041641235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.01,0.04553664207458496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.01,0.05972415924072265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.01,0.07190976142883301
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.01,0.0930508804321289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.01,0.11770303726196289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.01,0.16617536544799805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.01,0.21528255462646487
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,16,1,power_law_1.2,0.04325568199157715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,16,1,power_law_1.2,0.05073728084564209
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,16,1,power_law_1.2,0.052239999771118165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,16,1,power_law_1.2,0.062482562065124515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,16,1,power_law_1.2,0.08870271682739259
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,16,1,power_law_1.2,0.09086336135864258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,16,1,power_law_1.2,0.09208703994750976
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,16,1,power_law_1.2,0.09263808250427245
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,16,1,power_law_1.2,0.09389760017395019
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,16,1,power_law_1.2,0.09478143692016602
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,16,1,power_law_1.2,0.09786175727844239
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,16,1,power_law_1.2,0.09896063804626465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,16,1,power_law_1.2,0.10152640342712402
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,16,1,power_law_1.2,0.10555328369140624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,16,1,power_law_1.2,0.10885503768920898
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,16,1,power_law_1.2,0.11850815773010255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,16,1,power_law_1.2,0.1198624038696289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.01,0.3115904045104981
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,16,1,power_law_1.2,0.135034875869751
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,16,1,power_law_1.2,0.17037311553955078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,16,1,power_law_1.2,0.21988479614257814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,16,1,power_law_1.2,0.2597574424743653
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,16,1,power_law_1.2,0.355278091430664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,16,1,power_law_1.2,0.4451007843017578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.01,0.40831871032714845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,16,1,power_law_1.2,0.6330495834350586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.01,0.505656967163086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,16,1,power_law_1.2,0.8212691497802733
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,16,1,power_law_1.2,0.032695040702819825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,16,1,power_law_1.2,0.035360639095306394
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,16,1,power_law_1.2,0.03700992107391358
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,16,1,power_law_1.2,0.0393286395072937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,16,1,power_law_1.2,0.04228864192962646
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,16,1,power_law_1.2,0.043575038909912114
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,16,1,power_law_1.2,0.04403071880340576
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,16,1,power_law_1.2,0.0452019214630127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,16,1,power_law_1.2,0.04556735992431641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,16,1,power_law_1.2,0.04597695827484131
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,16,1,power_law_1.2,0.04715263843536377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,16,1,power_law_1.2,0.04888448238372803
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,16,1,power_law_1.2,0.04967872142791748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,16,1,power_law_1.2,0.055974397659301754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,16,1,power_law_1.2,0.06072192192077637
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,16,1,power_law_1.2,0.06395648002624513
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,16,1,power_law_1.2,0.06256000041961671
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,16,1,power_law_1.2,0.06938111782073975
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,16,1,power_law_1.2,0.0851699161529541
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,16,1,power_law_1.2,0.10646080017089843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,16,1,power_law_1.2,1.1995999908447266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,16,1,power_law_1.2,0.12022336006164551
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,16,1,power_law_1.2,0.16525632858276368
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,16,1,power_law_1.2,0.21154687881469725
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,16,1,power_law_1.2,0.2901932716369629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,16,1,power_law_1.2,1.5770944213867186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,16,1,power_law_1.2,0.3801785659790039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,16,1,power_law_1.2,1.9513836669921876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,16,1,power_law_1.2,0.018068480491638183
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,16,1,power_law_1.2,0.018506239652633667
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,16,1,power_law_1.2,0.018897279500961303
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,16,1,power_law_1.2,0.02052992105484009
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,16,1,power_law_1.2,0.02058176040649414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,16,1,power_law_1.2,0.020592000484466553
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,16,1,power_law_1.2,0.02079807996749878
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,16,1,power_law_1.2,0.0207859206199646
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,16,1,power_law_1.2,0.5456576156616211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,16,1,power_law_1.2,0.020952959060668946
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,16,1,power_law_1.2,0.02100735902786255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,16,1,power_law_1.2,0.022405118942260743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,16,1,power_law_1.2,0.02365119934082031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,16,1,power_law_1.2,0.025783040523529054
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,16,1,power_law_1.2,0.02878976106643677
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,16,1,power_law_1.2,0.03265856027603149
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,16,1,power_law_1.2,0.020956161022186278
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,16,1,power_law_1.2,0.03490816116333008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,16,1,power_law_1.2,0.713987808227539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,16,1,power_law_1.2,0.041208319664001465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,16,1,power_law_1.2,0.8714969635009766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,16,1,power_law_1.2,0.047621760368347164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,16,1,power_law_1.2,0.09631296157836913
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,16,1,power_law_1.2,0.06015808105468749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,16,1,power_law_1.2,0.07294400215148926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,16,1,power_law_1.2,0.22241088867187503
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,16,1,power_law_1.2,0.12187007904052734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,16,1,power_law_1.2,0.17161407470703124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.01,0.06081471920013428
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.01,0.06881792068481445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.01,0.07342400074005126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.01,0.08619456291198731
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.01,0.12227007865905762
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.01,0.12401856422424315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.01,0.12500032424926757
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.01,0.125797119140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.01,0.1268889617919922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.01,0.12853376388549803
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.01,0.13331007957458496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.01,0.13942336082458495
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.01,0.1443008041381836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.01,0.14770751953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.01,0.15110527992248535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.01,0.15940671920776367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.01,0.16769279479980467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.01,0.20829824447631834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.01,0.27444992065429685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.01,0.3350969696044922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.01,0.4058272171020508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.01,0.558149757385254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.01,0.7117791748046874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.01,1.0212844848632812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,16,1,power_law_1.2,0.32134208679199217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.01,1.3304025268554687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.01,1.9516076660156252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,16,1,power_law_1.2,0.5204998397827149
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.01,0.04176959991455078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.01,0.04809408187866211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.01,0.04973696231842041
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.01,0.06033152103424072
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.01,2.5714257812500003
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,16,1,power_law_1.2,0.42483776092529296
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.01,0.06629248142242432
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.01,0.0669478416442871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.01,0.06864895820617675
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.01,0.06845632076263428
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.01,0.06972864151000976
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.01,0.07202112197875976
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.01,0.07305791854858398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.01,0.07039296150207519
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.01,0.07504064083099365
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.01,0.07698624134063721
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.01,0.07924032211303711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.01,0.08231679916381836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.01,0.08322239875793457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.01,0.09409407615661622
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.01,0.12572671890258788
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.01,0.14356160163879395
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.01,0.1662495994567871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.01,0.23364288330078126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.01,0.29669120788574216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.01,0.4244819259643554
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.01,0.5434828948974609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,16,1,power_law_1.01,3.224268798828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.01,0.7925548553466797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.01,1.0272447967529297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.01,0.018993279933929443
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.01,0.028026239871978758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.01,0.023274879455566406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.01,0.029728639125823974
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.01,0.030412158966064452
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.01,0.030376319885253904
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.01,0.029155199527740476
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.01,0.02948096036911011
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.01,0.031006081104278566
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.01,0.03160831928253174
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,16,1,power_law_1.01,1.2855084228515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.01,0.023866240978240964
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.01,0.031571838855743405
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.01,0.03355648040771485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.01,0.03930943965911865
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.01,0.04104959964752197
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.01,0.04196544170379639
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.01,0.04652927875518799
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.01,0.05415359973907471
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.01,0.06472959995269775
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.01,0.07978047847747802
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.01,0.09429823875427246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.01,0.12986175537109373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.01,0.1640108871459961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.01,0.23021888732910156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.01,0.2981977653503418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,16,1,power_law_1.2,0.06071680068969727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,16,1,power_law_1.2,0.06793471813201904
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,16,1,power_law_1.2,0.07120960235595704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,16,1,power_law_1.2,0.08640128135681152
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,16,1,power_law_1.2,0.12214143753051758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,16,1,power_law_1.2,0.12479167938232423
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,16,1,power_law_1.2,0.12592000007629395
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,16,1,power_law_1.2,0.125927677154541
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,16,1,power_law_1.2,0.1274777603149414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,16,1,power_law_1.2,0.12854463577270508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,16,1,power_law_1.2,0.13385024070739746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,16,1,power_law_1.2,0.139748477935791
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,16,1,power_law_1.2,0.1431001567840576
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,16,1,power_law_1.2,0.1476364803314209
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,16,1,power_law_1.2,0.1515065574645996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,16,1,power_law_1.2,0.1604863929748535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,16,1,power_law_1.2,0.1674291229248047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,16,1,power_law_1.2,0.20937536239624022
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,16,1,power_law_1.2,0.27711872100830076
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.01,0.44026367187500004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,16,1,power_law_1.2,0.3361439895629883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,16,1,power_law_1.2,0.4072915267944336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,16,1,power_law_1.2,0.563309440612793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,16,1,power_law_1.2,0.7141120147705078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.01,0.5737311935424805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,16,1,power_law_1.2,1.0212191772460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.01,0.7110578918457031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,16,1,power_law_1.2,1.329246063232422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,16,1,power_law_1.2,0.040016641616821294
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,16,1,power_law_1.2,0.04814591884613037
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,16,1,power_law_1.2,0.04931839942932129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,16,1,power_law_1.2,0.061514878273010255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,16,1,power_law_1.2,0.06552576065063477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,16,1,power_law_1.2,0.06779967784881592
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,16,1,power_law_1.2,0.06892608165740967
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,16,1,power_law_1.2,0.06863935947418212
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,16,1,power_law_1.2,0.06999551773071289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,16,1,power_law_1.2,0.07048768043518067
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,16,1,power_law_1.2,0.07175551891326905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,16,1,power_law_1.2,0.07379712104797363
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,16,1,power_law_1.2,0.07504831790924073
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,16,1,power_law_1.2,0.0779852819442749
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,16,1,power_law_1.2,0.08027647972106934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,16,1,power_law_1.2,0.08292863845825196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,16,1,power_law_1.2,0.08644160270690918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,16,1,power_law_1.2,0.09519935607910156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,16,1,power_law_1.2,0.1251372814178467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,16,1,power_law_1.2,0.1447417640686035
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,16,1,power_law_1.2,0.16777023315429687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,16,1,power_law_1.2,1.9481829833984374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,16,1,power_law_1.2,0.2342899131774902
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,16,1,power_law_1.2,0.2960019111633301
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,16,1,power_law_1.2,2.56598388671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,16,1,power_law_1.2,0.42908287048339844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,16,1,power_law_1.2,0.5386943817138672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,16,1,power_law_1.2,3.3183117675781246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,16,1,power_law_1.2,0.019334399700164796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,16,1,power_law_1.2,0.02279360055923462
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,16,1,power_law_1.2,0.02350464105606079
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,16,1,power_law_1.2,0.027518720626831056
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,16,1,power_law_1.2,0.027854080200195315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,16,1,power_law_1.2,0.029466240406036376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,16,1,power_law_1.2,0.02982975959777832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,16,1,power_law_1.2,0.029154560565948485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,16,1,power_law_1.2,0.030418560504913327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,16,1,power_law_1.2,0.03071232080459595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,16,1,power_law_1.2,0.03138880014419555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,16,1,power_law_1.2,0.03204031944274902
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,16,1,power_law_1.2,0.03380032062530518
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,16,1,power_law_1.2,0.03913536071777344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,16,1,power_law_1.2,0.042675199508666994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,16,1,power_law_1.2,0.04394495964050293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,16,1,power_law_1.2,0.7908882904052734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,16,1,power_law_1.2,0.04676032066345215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,16,1,power_law_1.2,0.05431168079376221
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,16,1,power_law_1.2,1.0382848358154297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,16,1,power_law_1.2,0.06536896228790283
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,16,1,power_law_1.2,0.08053312301635743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,16,1,power_law_1.2,0.09724543571472169
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,16,1,power_law_1.2,1.2730368041992188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,16,1,power_law_1.2,0.13302016258239746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,16,1,power_law_1.2,0.16831296920776367
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,16,1,power_law_1.2,0.23986879348754883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,16,1,power_law_1.2,0.3097478485107422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.01,0.045573759078979495
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.01,0.061910400390624995
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.01,0.0833619213104248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.01,0.10768447875976564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.01,0.13646143913269043
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.01,0.1770310401916504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.01,0.22681856155395508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.01,0.2368332862854004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.01,0.24086912155151366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.01,0.2454355239868164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.01,0.25900863647460937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.01,0.2877471923828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.01,0.3059449577331543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.01,0.32385406494140623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.01,0.3518060684204102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.01,0.3591584014892578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.01,0.3801043319702148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.01,0.42514816284179685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.01,0.5068518447875976
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.01,0.5692083358764648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.01,0.6443116760253906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.01,0.8104115295410157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.01,0.9505548858642578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.01,1.282782745361328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.01,1.615614013671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.01,2.272679748535156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.01,2.9321694946289063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,16,1,power_law_1.2,0.45003391265869136
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,16,1,power_law_1.01,3.5998593139648434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.01,0.04373568058013916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.01,0.050919680595397955
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.01,0.10171392440795898
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.01,0.07025983810424805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.01,0.1413369560241699
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.01,0.1981862449645996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.01,0.24940544128417969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.01,0.25610240936279294
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.01,0.25986303329467775
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.01,0.2637126350402832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.01,0.2726041603088379
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.01,0.2798080062866211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.01,0.2850892829895019
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.01,0.29121599197387693
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.01,0.32046913146972655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.01,0.3248518371582031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.01,0.3377094268798828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.01,0.3596160125732422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.01,0.38626880645751954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.01,0.43509185791015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.01,0.4987257766723633
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.01,0.6017958450317382
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.01,0.702775650024414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.01,0.9208812713623047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.01,1.1527206420898437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.01,1.596050567626953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.01,2.0212185668945315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,16,1,power_law_1.01,2.469432373046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,16,1,power_law_1.2,0.597086067199707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.01,0.06487679958343506
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.01,0.09445952415466309
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.01,0.06812160015106201
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.01,0.11507712364196779
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.01,0.1332793617248535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.01,0.1658515167236328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.01,0.17291135787963868
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.01,0.08069567680358887
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.01,0.17678016662597656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.01,0.1801900863647461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.01,0.18676992416381838
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,16,1,power_law_1.2,0.7384870147705078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.01,0.19342975616455077
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.01,0.19921344757080078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.01,0.20686784744262696
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.01,0.22187583923339843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.01,0.23994623184204103
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.01,0.21619455337524415
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.01,0.2754380798339844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.01,0.3150662422180176
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.01,0.39882049560546873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.01,0.4805267333984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.01,0.6459986877441406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.01,0.8139891052246094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.01,1.2207008361816407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.01,1.6513337707519533
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,16,1,power_law_1.01,2.471482849121094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,16,1,power_law_1.2,0.04313280105590821
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,16,1,power_law_1.2,0.06120448112487793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,16,1,power_law_1.2,0.10389439582824707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,16,1,power_law_1.01,3.133982849121094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,16,1,power_law_1.2,0.1305427169799805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,16,1,power_law_1.2,0.08329855918884277
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,16,1,power_law_1.2,0.17616191864013672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,16,1,power_law_1.2,0.22309696197509768
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,16,1,power_law_1.2,0.235731201171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,16,1,power_law_1.01,3.8899993896484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,16,1,power_law_1.2,0.23842432022094728
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,16,1,power_law_1.2,0.24707199096679688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,16,1,power_law_1.2,0.26141504287719725
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,16,1,power_law_1.2,0.3080691146850586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,16,1,power_law_1.2,0.29158016204833986
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,16,1,power_law_1.2,0.33253631591796873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,16,1,power_law_1.2,0.35588417053222654
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,16,1,power_law_1.2,0.36780609130859376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,16,1,power_law_1.2,0.38725120544433594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,16,1,power_law_1.2,0.4495635223388672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,16,1,power_law_1.2,0.517507209777832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,16,1,power_law_1.2,0.5777107238769531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,16,1,power_law_1.2,0.6664569854736329
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,16,1,power_law_1.2,0.8237619018554687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,16,1,power_law_1.2,0.9644300842285156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,16,1,power_law_1.2,1.2903730773925781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,16,1,power_law_1.2,1.629416961669922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,16,1,power_law_1.2,2.287018280029297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,16,1,power_law_1.2,0.050506877899169925
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,16,1,power_law_1.2,0.044360961914062504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,16,1,power_law_1.2,0.07158080101013184
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,16,1,power_law_1.2,2.9560986328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,16,1,power_law_1.2,0.13484928131103516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,16,1,power_law_1.2,0.18889087677001953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,16,1,power_law_1.2,0.0966323184967041
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,16,1,power_law_1.2,0.24670143127441407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,16,1,power_law_1.2,3.6275341796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,16,1,power_law_1.2,0.2535935974121094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,16,1,power_law_1.2,0.257891845703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,16,1,power_law_1.2,0.2639801597595215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,16,1,power_law_1.2,0.27000448226928714
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,16,1,power_law_1.2,0.278929271697998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,16,1,power_law_1.2,0.28450176239013675
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,16,1,power_law_1.2,0.2952582359313965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,16,1,power_law_1.2,0.32268672943115234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,16,1,power_law_1.2,0.33332286834716796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,16,1,power_law_1.2,0.37324607849121094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,16,1,power_law_1.2,0.3983712005615234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,16,1,power_law_1.2,0.45599998474121095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,16,1,power_law_1.2,0.5161203384399414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,16,1,power_law_1.2,0.34302719116210934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,16,1,power_law_1.2,0.6198009490966797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,16,1,power_law_1.2,0.7408608245849609
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,16,1,power_law_1.2,0.9337689971923828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,16,1,power_law_1.2,1.1663168334960938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,16,1,power_law_1.2,1.611287078857422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,16,1,power_law_1.2,0.06482367992401124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,16,1,power_law_1.2,2.0588896179199216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,16,1,power_law_1.2,0.06850111961364747
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,16,1,power_law_1.2,0.07923136234283448
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,16,1,power_law_1.2,2.48321533203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,16,1,power_law_1.2,0.0920576000213623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,16,1,power_law_1.2,0.11007871627807617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,16,1,power_law_1.2,0.1292249584197998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,16,1,power_law_1.2,0.16422975540161133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,16,1,power_law_1.2,0.17169343948364257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,16,1,power_law_1.2,0.1756800079345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,16,1,power_law_1.2,0.18154752731323243
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,16,1,power_law_1.2,0.18799808502197266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,16,1,power_law_1.2,0.19585279464721678
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,16,1,power_law_1.2,0.2142835235595703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,16,1,power_law_1.2,0.20215423583984374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,16,1,power_law_1.2,0.227260799407959
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,16,1,power_law_1.2,0.23838016510009769
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,16,1,power_law_1.2,0.2637676811218262
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,16,1,power_law_1.2,0.3781465530395508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,16,1,power_law_1.2,0.5202751922607421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,16,1,power_law_1.2,0.32014270782470705
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,16,1,power_law_1.2,0.6427423858642578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,16,1,power_law_1.2,0.946948471069336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,16,1,power_law_1.2,1.2613254547119142
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,16,1,power_law_1.2,1.8682144165039063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.01,0.03128511905670166
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.01,0.04699264049530029
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.01,0.06352960109710694
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.01,0.07443391799926757
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.01,0.09086272239685059
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.01,0.09209792137145996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.01,0.09510272026062012
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.01,0.09819071769714356
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.01,0.10119359970092774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.01,0.10672896385192872
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.01,0.13512895584106446
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.01,0.13812607765197754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.01,0.1428985595703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.01,0.14615296363830566
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.01,0.1495404815673828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.01,0.15861696243286133
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.01,0.17687488555908204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.01,0.19269632339477538
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.01,0.22897151947021482
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.01,0.2709152030944824
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.01,0.3799929428100586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.01,0.4554617691040039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.01,0.6405382537841797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.01,0.8094175720214845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.01,1.1608262634277344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.01,1.5109484863281248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,16,1,power_law_1.01,1.868153533935547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.01,0.03408128023147583
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.01,0.03943039894104004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.01,0.05159423828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.01,0.06073791980743408
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.01,0.07539199829101563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,16,1,power_law_1.2,2.5836517333984377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.01,0.0950380802154541
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.01,0.09803327560424804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.01,0.10179007530212403
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.01,0.10059136390686034
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.01,0.10217344284057617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.01,0.10393792152404786
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.01,0.1180998420715332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.01,0.11902527809143067
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.01,0.12529472351074217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.01,0.12463487625122069
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.01,0.12886143684387208
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.01,0.13714367866516114
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.01,0.14999679565429686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.01,0.17376895904541018
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.01,0.20539392471313475
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.01,0.23942399978637696
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.01,0.30090368270874024
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.01,0.37418495178222655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.01,0.5112838363647461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.01,0.6351500701904297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.01,0.9056166076660157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.01,1.1792940521240234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,16,1,power_law_1.01,1.443066864013672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.01,0.04273727893829345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.01,0.05045504093170165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.01,0.04524799823760987
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.01,0.05564352035522461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.01,0.061515522003173825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.01,0.07739456176757813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.01,0.0805510425567627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.01,0.08311167716979981
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.01,0.08396544456481933
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.01,0.08553024291992187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.01,0.08848383903503418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.01,0.09159999847412109
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.01,0.09236096382141114
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.01,0.09964032173156738
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.01,0.10887616157531739
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.01,0.11252032279968263
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.01,0.14469696044921876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.01,0.12234496116638183
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.01,0.17230207443237305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.01,0.22497087478637695
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.01,0.2746188735961914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.01,0.38480575561523434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.01,0.4972851181030274
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.01,0.73565185546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.01,0.9903040313720703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.01,1.4880326843261718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,16,1,power_law_1.2,0.033724160194396974
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,16,1,power_law_1.2,0.05141568183898926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,16,1,power_law_1.2,0.05579328060150146
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,16,1,power_law_1.2,0.06191423892974853
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,16,1,power_law_1.2,0.07552383899688721
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,16,1,power_law_1.2,0.09078656196594238
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,16,1,power_law_1.2,0.09354751586914063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,16,1,power_law_1.2,0.0998367977142334
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,16,1,power_law_1.2,0.09635199546813965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,16,1,power_law_1.2,0.1024294376373291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,16,1,power_law_1.2,0.10743167877197266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,16,1,power_law_1.2,0.13634431838989258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,16,1,power_law_1.2,0.13921792030334473
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,16,1,power_law_1.2,0.14420543670654298
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,16,1,power_law_1.2,0.14672767639160156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,16,1,power_law_1.2,0.15073280334472655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,16,1,power_law_1.2,0.1596294403076172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,16,1,power_law_1.2,0.1777516746520996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,16,1,power_law_1.2,0.19415679931640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,16,1,power_law_1.2,0.23042240142822265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,16,1,power_law_1.2,0.2760620880126953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,power_law_1.01,1.9573452758789063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,16,1,power_law_1.2,0.3814028930664063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,16,1,power_law_1.01,2.534686737060547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,16,1,power_law_1.2,0.4569023895263672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,16,1,power_law_1.2,0.6406406402587891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,16,1,power_law_1.2,0.8124992370605468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,16,1,power_law_1.2,0.035109119415283205
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,16,1,power_law_1.2,0.03988159894943237
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,16,1,power_law_1.2,0.05104832172393799
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,16,1,power_law_1.2,0.06076032161712647
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,16,1,power_law_1.2,0.0759168004989624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,16,1,power_law_1.2,0.10080320358276366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,16,1,power_law_1.2,0.09775808334350586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,16,1,power_law_1.2,0.09951423645019532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,16,1,power_law_1.2,0.10240511894226074
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,16,1,power_law_1.2,0.10265279769897462
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,16,1,power_law_1.2,0.10568256378173828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,16,1,power_law_1.2,0.11921152114868164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,16,1,power_law_1.2,0.12150400161743165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,16,1,power_law_1.2,0.1258937644958496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,16,1,power_law_1.2,0.12758784294128417
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,16,1,power_law_1.2,0.1315833568572998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,16,1,power_law_1.2,0.13815808296203613
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,16,1,power_law_1.2,0.1536467170715332
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,16,1,power_law_1.2,0.17911808013916014
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,16,1,power_law_1.2,0.2118899154663086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,16,1,power_law_1.2,0.24269376754760744
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,16,1,power_law_1.2,0.30281728744506836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,16,1,power_law_1.2,0.37996673583984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,16,1,power_law_1.2,1.1672531127929688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,16,1,power_law_1.2,0.5156409454345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,16,1,power_law_1.2,0.6369952011108398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,16,1,power_law_1.2,1.5153517150878906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,16,1,power_law_1.2,0.9101920318603515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,1536,8,128,16,1,power_law_1.2,0.043181438446044926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,1536,8,128,16,1,power_law_1.2,0.04606527805328369
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,1536,8,128,16,1,power_law_1.2,0.04994624137878418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,1536,8,128,16,1,power_law_1.2,0.05411968231201172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,1536,8,128,16,1,power_law_1.2,0.06008639812469483
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,1536,8,128,16,1,power_law_1.2,0.07797887802124023
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,16,1,power_law_1.2,1.8714047241210938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,1536,8,128,16,1,power_law_1.2,0.08116031646728515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,1536,8,128,16,1,power_law_1.2,0.08260543823242188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,1536,8,128,16,1,power_law_1.2,0.08578816413879395
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,1536,8,128,16,1,power_law_1.2,0.08599360466003418
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,1536,8,128,16,1,power_law_1.2,0.08983872413635255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,1536,8,128,16,1,power_law_1.2,0.09257792472839356
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,1536,8,128,16,1,power_law_1.2,0.09624192237854004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,1536,8,128,16,1,power_law_1.2,0.1089907169342041
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,1536,8,128,16,1,power_law_1.2,0.12284095764160156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,16,1,power_law_1.2,1.177717742919922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,1536,8,128,16,1,power_law_1.2,0.11553215980529785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,1536,8,128,16,1,power_law_1.2,0.13779264450073242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,1536,8,128,16,1,power_law_1.2,0.17382976531982422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,1536,8,128,16,1,power_law_1.2,0.20936447143554687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,1536,8,128,16,1,power_law_1.2,0.28812864303588864
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,1536,8,128,16,1,power_law_1.2,0.3657689666748047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,1536,8,128,16,1,power_law_1.2,0.5355654525756836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,16,1,power_law_1.2,1.4456031799316407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,1536,8,128,16,1,power_law_1.2,0.7077536010742188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,1536,8,128,16,1,power_law_1.2,1.0484012603759765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,1536,8,128,16,1,power_law_1.2,1.3843475341796876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.01,0.04277632236480713
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.01,0.06880512237548828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.01,0.08450112342834473
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.01,0.10419455528259278
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.01,0.12887999534606934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.01,0.18468544006347656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.01,0.18233535766601563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.01,0.1903590393066406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.01,0.19756160736083986
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.01,0.2021311950683594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.01,0.21700416564941408
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.01,0.23870719909667967
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.01,0.24419456481933594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.01,0.2530201530456543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.01,0.25666624069213867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.01,0.2613715171813965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.01,0.27899391174316407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.01,0.32178688049316406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.01,0.38451072692871097
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.01,0.45434944152832035
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.01,0.5259430313110351
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.01,0.6983372497558593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.01,0.8442655944824219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.01,1.167044448852539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.01,1.489624328613281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,1536,8,128,16,1,power_law_1.2,2.0753196716308593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.01,2.134169616699219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.01,2.7811859130859373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.01,0.04599936008453369
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,16,1,power_law_1.01,3.435086669921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.01,0.053946881294250494
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.01,0.07227007865905762
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.01,0.09483648300170898
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.01,0.1260371208190918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.01,0.18746623992919922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.01,0.1827340888977051
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.01,0.18672639846801758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.01,0.191331844329834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.01,0.19311935424804688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.01,0.19777088165283202
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.01,0.2039833641052246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.01,0.23458688735961913
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.01,0.249800968170166
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.01,0.24957248687744144
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.01,0.24078336715698243
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.01,0.27224767684936524
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.01,0.29549760818481446
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.01,0.31525888442993166
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.01,0.3631507110595703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.01,0.4365926361083984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.01,0.5478406524658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.01,0.6507443237304688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.01,0.89368896484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.01,1.1159852600097655
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.01,1.5785165405273438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.01,2.0258860778808594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,16,1,power_law_1.01,2.5132920837402346
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,1536,8,128,16,1,power_law_1.2,2.749688415527344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.01,0.049317121505737305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.01,0.07635583877563476
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.01,0.05445695877075195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.01,0.09044095993041992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.01,0.12391807556152343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.01,0.12791808128356935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.01,0.134203519821167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.01,0.13723072052001953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.01,0.1404595184326172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.01,0.14360320091247558
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.01,0.15115776062011718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.01,0.15022272109985352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.01,0.1634489631652832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.01,0.16718847274780274
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.01,0.17191104888916015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.01,0.06538879871368408
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,1536,8,128,16,1,power_law_1.2,3.430420532226562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.01,0.18836351394653322
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.01,0.22449855804443358
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.01,0.25356927871704105
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.01,0.327127685546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.01,0.40620990753173825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.01,0.5605510330200195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.01,0.7283372497558593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.01,1.0579513549804687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.01,1.4122918701171874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,16,1,power_law_1.2,0.04234432220458985
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,16,1,power_law_1.2,0.06900864124298096
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,16,1,power_law_1.2,0.08331135749816895
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,16,1,power_law_1.2,0.0998259162902832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.01,2.124156188964844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,16,1,power_law_1.2,0.12452799797058105
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,16,1,power_law_1.2,0.18496000289916992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.01,2.8510644531250002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,16,1,power_law_1.2,0.19390207290649414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,16,1,power_law_1.2,0.2009369659423828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,16,1,power_law_1.2,0.21138303756713867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,16,1,power_law_1.2,0.22140672683715817
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,16,1,power_law_1.2,0.253436164855957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,16,1,power_law_1.2,0.18037376403808594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,16,1,power_law_1.2,0.24614336013793947
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,16,1,power_law_1.2,0.25259584426879883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,16,1,power_law_1.2,0.2585343933105469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,16,1,power_law_1.2,0.2653696060180664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,16,1,power_law_1.2,0.28503744125366215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,16,1,power_law_1.2,0.3318214416503906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,16,1,power_law_1.2,0.40091838836669924
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,16,1,power_law_1.2,0.45759105682373047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,16,1,power_law_1.2,0.5326547241210937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,16,1,power_law_1.2,0.7001312255859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,16,1,power_law_1.2,0.8493299102783203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,16,1,power_law_1.2,1.1711347198486328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,16,1,power_law_1.2,1.4931494140625001
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,16,1,power_law_1.2,2.14391357421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,16,1,power_law_1.2,0.045574398040771485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,16,1,power_law_1.2,0.0576639986038208
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,16,1,power_law_1.2,0.071976318359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,16,1,power_law_1.2,0.08997119903564453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,16,1,power_law_1.2,0.12134464263916016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,2560,8,160,16,1,power_law_1.01,3.5941836547851564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,16,1,power_law_1.2,0.18622848510742188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,16,1,power_law_1.2,0.18110528945922852
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,16,1,power_law_1.2,0.18806848526000977
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,16,1,power_law_1.2,0.19205951690673828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,16,1,power_law_1.2,0.19569280624389648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,16,1,power_law_1.2,0.19989376068115233
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,16,1,power_law_1.2,0.20591615676879882
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,16,1,power_law_1.2,0.2500876808166504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,16,1,power_law_1.2,0.25116735458374023
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,16,1,power_law_1.2,0.26174911499023434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,16,1,power_law_1.2,0.2693433570861816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,16,1,power_law_1.2,0.28057983398437497
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,16,1,power_law_1.2,0.29885568618774416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,16,1,power_law_1.2,0.3215353775024414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,16,1,power_law_1.2,0.3727180862426758
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,16,1,power_law_1.2,0.44026622772216795
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,16,1,power_law_1.2,0.5634732818603515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,16,1,power_law_1.2,0.6616774749755859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,16,1,power_law_1.2,0.8998489379882812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,16,1,power_law_1.2,2.790757751464844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,16,1,power_law_1.2,1.121694107055664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,16,1,power_law_1.2,1.588153533935547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,2560,8,160,16,1,power_law_1.2,0.04969408035278321
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,2560,8,160,16,1,power_law_1.2,0.055160961151123046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,2560,8,160,16,1,power_law_1.2,0.06276671886444092
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,16,1,power_law_1.2,3.4418515014648436
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,2560,8,160,16,1,power_law_1.2,0.07202112197875976
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,2560,8,160,16,1,power_law_1.2,0.08500351905822753
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,2560,8,160,16,1,power_law_1.2,0.12092351913452148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,2560,8,160,16,1,power_law_1.2,0.1278048038482666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,2560,8,160,16,1,power_law_1.2,0.13394944190979002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,2560,8,160,16,1,power_law_1.2,0.13966527938842774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,2560,8,160,16,1,power_law_1.2,0.1425011157989502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,2560,8,160,16,1,power_law_1.2,0.14999360084533692
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,2560,8,160,16,1,power_law_1.2,0.15218239784240722
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,2560,8,160,16,1,power_law_1.2,0.1609503936767578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,2560,8,160,16,1,power_law_1.2,0.16744447708129884
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,16,1,power_law_1.2,2.040533142089844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,16,1,power_law_1.2,2.5047206115722656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,2560,8,160,16,1,power_law_1.2,0.18139520645141602
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,2560,8,160,16,1,power_law_1.2,0.1933420753479004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,2560,8,160,16,1,power_law_1.2,0.2127756881713867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,2560,8,160,16,1,power_law_1.2,0.26135360717773437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,2560,8,160,16,1,power_law_1.2,0.31074176788330077
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,2560,8,160,16,1,power_law_1.2,0.42991104125976565
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,2560,8,160,16,1,power_law_1.2,0.5515827178955078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,2560,8,160,16,1,power_law_1.2,0.8255661010742188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,2560,8,160,16,1,power_law_1.2,1.1098355102539064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,2560,8,160,16,1,power_law_1.2,1.7199098205566408
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,2560,8,160,16,1,power_law_1.2,2.342640075683594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.01,0.04407423973083496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.01,0.07185344219207764
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.01,0.08746175765991211
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.01,0.1134995174407959
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.01,0.15969344139099123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.01,0.2252851104736328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.01,0.25138240814208984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.01,0.32763584136962887
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.01,0.34507839202880863
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.01,0.3616384124755859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.01,0.37125953674316403
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.01,0.38075584411621094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.01,0.39390655517578127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.01,0.4053452682495117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.01,0.4156198501586914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.01,0.44030654907226563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.01,0.5706630325317382
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.01,0.5943059158325196
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.01,0.6278803253173828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.01,0.7500704193115235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.01,0.8141824340820312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.01,0.9406233978271483
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,2560,8,160,16,1,power_law_1.2,3.597984008789062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.01,1.1073267364501953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.01,1.449920654296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.01,1.7767333984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.01,2.4652307128906252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.01,3.139177551269531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.01,0.04761216163635254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.01,0.053408641815185544
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.01,0.07114240169525146
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.01,0.10677120208740234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.01,0.1639360046386719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.01,0.28553152084350586
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,16,1,power_law_1.01,3.8277081298828124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.01,0.32782913208007813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.01,0.3389344024658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.01,0.3497510528564453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.01,0.35838142395019534
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.01,0.36562049865722657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.01,0.37759552001953123
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.01,0.395074577331543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.01,0.3974886322021484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.01,0.4605689620971679
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.01,0.41441600799560546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.01,0.48307968139648433
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.01,0.5075193786621094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.01,0.5613043212890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.01,0.6140153503417969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.01,0.72853759765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.01,0.8469625854492188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,2560,8,160,16,1,power_law_1.2,4.852880554199219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.01,1.0836428833007812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.01,1.3066329956054688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.01,1.7675059509277342
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.01,2.2329913330078126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,16,1,power_law_1.01,2.7010098266601563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.01,0.0804089641571045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.01,0.08319552421569824
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.01,0.08040255546569824
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.01,0.08593215942382812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.01,0.09141887664794922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.01,0.08081343650817871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.01,0.09799424171447754
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.01,0.10337984085083007
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.01,0.10960703849792482
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.01,0.11647744178771972
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.01,0.12829055786132812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.01,0.15209152221679687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.01,0.1672537612915039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.01,0.21357120513916014
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.01,0.2741375923156738
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.01,0.31254207611083984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.01,0.39459518432617186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.01,0.5564236831665039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.01,0.7184467315673828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.01,1.0417638397216797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.01,1.3608953857421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.01,2.016174774169922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.01,2.686316833496094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,16,1,power_law_1.2,0.046057600975036625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,16,1,power_law_1.2,0.06498303890228271
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,16,1,power_law_1.2,0.08693375587463378
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,16,1,power_law_1.2,0.11152192115783692
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.01,3.9675173950195317
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,16,1,power_law_1.2,0.14838848114013672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,16,1,power_law_1.2,0.20585216522216795
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,16,1,power_law_1.2,0.23582336425781253
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,16,1,power_law_1.2,0.3266342544555664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,16,1,power_law_1.2,0.3377587127685547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,16,1,power_law_1.2,0.3562483215332031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,16,1,power_law_1.2,0.3654886245727539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,16,1,power_law_1.2,0.38211200714111326
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,16,1,power_law_1.2,0.3918003082275391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,16,1,power_law_1.2,0.40364608764648435
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,16,1,power_law_1.2,0.44604545593261713
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,16,1,power_law_1.2,0.42051647186279295
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,16,1,power_law_1.2,0.5678815841674805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,16,1,power_law_1.2,0.5948844909667969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,16,1,power_law_1.2,0.6334982299804688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,16,1,power_law_1.2,0.7594335937500001
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,16,1,power_law_1.2,0.820782699584961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,16,1,power_law_1.2,0.9517203521728514
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,16,1,power_law_1.2,1.1217420959472657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,16,1,power_law_1.2,1.4614898681640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,16,1,power_law_1.2,1.801966094970703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,16,1,power_law_1.2,2.480435791015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,16,1,power_law_1.2,3.1657965087890623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,16,1,power_law_1.2,0.046779518127441404
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,16,1,power_law_1.2,0.05256319999694824
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,16,1,power_law_1.2,0.07258304119110107
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,16,1,power_law_1.2,3.8582366943359374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,16,1,power_law_1.2,0.10479040145874023
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,16,1,power_law_1.2,0.1513849639892578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,16,1,power_law_1.2,0.22228351593017578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,16,1,power_law_1.2,0.27615999221801757
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,16,1,power_law_1.2,0.3226553726196289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,16,1,power_law_1.2,0.3332928085327148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,16,1,power_law_1.2,0.34060863494873045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,16,1,power_law_1.2,0.3512287902832031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,16,1,power_law_1.2,0.3627763366699219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,16,1,power_law_1.2,0.37517566680908204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,16,1,power_law_1.2,0.3915353775024414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,16,1,power_law_1.2,0.39919681549072267
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,16,1,power_law_1.2,0.4129657745361328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,16,1,power_law_1.2,0.4588966369628906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,16,1,power_law_1.2,0.4920870590209961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,16,1,power_law_1.2,0.5178387069702148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,16,1,power_law_1.2,0.5758713531494141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,16,1,power_law_1.2,0.6352441787719727
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,16,1,power_law_1.2,0.7668857574462891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,16,1,power_law_1.2,0.8756806182861329
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,16,1,power_law_1.2,1.1092838287353515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,16,1,power_law_1.2,1.343816375732422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,16,1,power_law_1.2,1.8005357360839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,16,1,power_law_1.2,2.2670022583007814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,16,1,power_law_1.2,2.7331661987304687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,16,1,power_law_1.2,0.0800063991546631
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,16,1,power_law_1.2,0.08091839790344238
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,16,1,power_law_1.2,0.08180031776428223
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,16,1,power_law_1.2,0.08486592292785644
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,16,1,power_law_1.2,0.09152064323425294
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,16,1,power_law_1.2,0.09728447914123535
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,16,1,power_law_1.2,0.1164352035522461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,16,1,power_law_1.2,0.12813055992126465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,16,1,power_law_1.2,0.15275903701782229
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,16,1,power_law_1.2,0.17312320709228515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,16,1,power_law_1.2,0.08003135681152343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,16,1,power_law_1.2,0.20846464157104494
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,16,1,power_law_1.2,0.27693567276000974
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,16,1,power_law_1.2,0.3122649574279785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,16,1,power_law_1.2,0.10285951614379882
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,16,1,power_law_1.2,0.10870400428771973
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,16,1,power_law_1.2,0.3978374481201172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,16,1,power_law_1.2,0.5549612808227539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,16,1,power_law_1.2,0.7156883239746094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,16,1,power_law_1.2,1.0401241302490234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,16,1,power_law_1.2,1.3695289611816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,16,1,power_law_1.2,2.0098757934570313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,16,1,power_law_1.2,2.6759521484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,16,1,power_law_1.2,3.9673202514648436
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,power_law_1.01,0.030035839080810544
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,power_law_1.01,0.03296704053878784
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.01,0.05310336112976074
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.01,0.06548480033874512
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.01,0.07952896118164063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.01,0.10771264076232909
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.01,0.11840576171874999
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.01,0.11506496429443361
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.01,0.1220582389831543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.01,0.12787072181701659
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.01,0.13533439636230468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.01,0.16284032821655275
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.01,0.17680192947387696
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.01,0.18709951400756836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.01,0.18767488479614258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.01,0.19298816680908204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.01,0.20128192901611328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.01,0.22220800399780272
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.01,0.24283903121948242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.01,0.29924415588378905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.01,0.3479840087890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.01,0.44955585479736326
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.01,0.5532614517211913
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.01,0.7426502227783203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.01,0.9530054473876953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.01,1.351345977783203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.01,1.7572064208984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,16,1,power_law_1.01,2.165679321289063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,16,1,power_law_1.2,0.03452863931655884
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,16,1,power_law_1.2,0.05270847797393798
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,16,1,power_law_1.2,0.028005759716033935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,16,1,power_law_1.2,0.06230207920074463
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,16,1,power_law_1.2,0.0748902416229248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,16,1,power_law_1.2,0.10870335578918458
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,16,1,power_law_1.2,0.11496064186096192
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,16,1,power_law_1.2,0.12338111877441407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,16,1,power_law_1.2,0.12045439720153808
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,16,1,power_law_1.2,0.12977279663085936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,16,1,power_law_1.2,0.1384812831878662
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,16,1,power_law_1.2,0.18140224456787107
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,16,1,power_law_1.2,0.18543039321899416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,16,1,power_law_1.2,0.18951616287231446
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,16,1,power_law_1.2,0.1920243263244629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,16,1,power_law_1.2,0.1959596824645996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,16,1,power_law_1.2,0.2065836715698242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,16,1,power_law_1.2,0.22520576477050783
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,16,1,power_law_1.2,0.24512832641601562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,16,1,power_law_1.2,0.31013504028320316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,16,1,power_law_1.2,0.3494911956787109
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,16,1,power_law_1.2,0.4526700973510742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,16,1,power_law_1.2,0.550179214477539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,16,1,power_law_1.2,0.7424550628662109
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,16,1,power_law_1.2,0.9542630767822265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,16,1,power_law_1.2,1.3545382690429686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,16,1,power_law_1.2,1.7648492431640626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.01,0.10923711776733398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.01,0.17650047302246094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.01,0.2779046440124512
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.01,0.40767166137695315
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.01,0.6446701049804687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.01,0.6855359649658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.01,0.07079808235168457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.01,0.7146559906005859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.01,0.7385126495361328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.01,0.8094931030273438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.01,0.8763577270507813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.01,0.7537184143066407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.01,0.9325363159179687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.01,1.0684339141845702
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.01,1.1578419494628907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.01,1.2907308959960937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,16,1,power_law_1.2,2.182213134765625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.01,1.5338925170898439
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.01,1.3581893920898438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.01,1.2260205078125002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.01,1.9635417175292968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.01,1.6595404052734373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.01,2.5664523315429686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.01,3.160418701171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.01,4.369627380371094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.01,5.59115234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.01,8.044532470703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.01,10.494776611328124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,16,1,power_law_1.01,12.948395996093751
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.01,0.3947238540649414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.01,0.2320454406738281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.01,0.5989209747314452
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.01,0.9966175842285157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.01,1.6913279724121093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.01,2.8793869018554688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.01,3.111036071777344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.01,3.1958804321289063
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.01,2.2511839294433593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.01,2.4244908142089843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.01,2.553833618164062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.01,2.7323461914062497
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.01,3.302933044433594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.01,2.720941467285156
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.01,2.945213928222656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.01,3.97393798828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.01,3.330768737792969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.01,3.855438842773437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.01,5.6500396728515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.01,5.398202514648437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.01,6.424608764648437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.01,6.6477783203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.01,10.81982421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.01,10.3186572265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.01,17.927210693359378
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,16,1,power_law_1.2,0.07269311904907226
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.01,17.881600341796876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,16,1,power_law_1.2,0.11373120307922364
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,16,1,power_law_1.2,0.17597951889038085
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,16,1,power_law_1.2,0.26183103561401366
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,16,1,power_law_1.2,0.6350860977172852
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,16,1,power_law_1.2,0.6928307342529296
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,16,1,power_law_1.2,0.3743244934082031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,16,1,power_law_1.2,0.7117209625244141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,16,1,power_law_1.2,0.7463724517822266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,16,1,power_law_1.2,0.7760031890869141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,16,1,power_law_1.2,0.8373049926757812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,16,1,power_law_1.2,0.9074400329589845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,16,1,power_law_1.2,0.9680307006835938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,16,1,power_law_1.2,1.1026457977294921
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,16,1,power_law_1.2,1.2113113403320312
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,16,1,power_law_1.2,1.3280493164062501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,16,1,power_law_1.2,1.5750962829589843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,16,1,power_law_1.2,1.264540786743164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,16,1,power_law_1.2,1.411751708984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,16,1,power_law_1.2,1.6947039794921874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,16,1,power_law_1.2,1.9946322631835938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,16,1,power_law_1.01,27.981723632812496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,16,1,power_law_1.2,2.5903668212890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.01,32.56592041015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,16,1,power_law_1.2,3.1891009521484373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,16,1,power_law_1.2,4.410397338867187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,16,1,power_law_1.2,5.62225341796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,16,1,power_law_1.2,0.2287615966796875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,16,1,power_law_1.2,0.35278526306152347
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,16,1,power_law_1.2,0.6475507354736327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,16,1,power_law_1.2,8.074017944335939
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,16,1,power_law_1.2,1.393621826171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,16,1,power_law_1.2,3.2207769775390624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,16,1,power_law_1.2,3.3961727905273436
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,16,1,power_law_1.2,1.2604544067382812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,16,1,power_law_1.2,3.5353695678710935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,16,1,power_law_1.2,3.4987603759765626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,16,1,power_law_1.2,3.7630874633789064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,16,1,power_law_1.2,3.7333343505859373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,16,1,power_law_1.2,3.9540914916992187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,16,1,power_law_1.2,10.534630126953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,16,1,power_law_1.2,3.8887680053710936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,16,1,power_law_1.2,2.9384506225585936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,16,1,power_law_1.2,4.254240112304688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,16,1,power_law_1.2,4.431524353027344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,16,1,power_law_1.2,3.2500595092773437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,16,1,power_law_1.2,5.5452728271484375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,16,1,power_law_1.2,5.889447631835937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,16,1,power_law_1.2,4.988125305175781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,16,1,power_law_1.2,7.662902221679687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,16,1,power_law_1.2,9.350948486328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,16,1,power_law_1.2,8.9300048828125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,16,1,power_law_1.2,12.997034912109376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,16,1,power_law_1.2,14.753238525390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,16,1,power_law_1.2,18.26457275390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.01,0.027587840557098387
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.01,0.03739775896072388
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.01,0.03901823997497559
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.01,0.0398304009437561
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.01,0.06263936042785645
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.01,0.06319359779357911
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.01,0.06466944217681884
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.01,0.06557695865631104
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.01,0.06489664077758789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.01,0.07138751983642579
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.01,0.07496895790100097
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.01,0.07887743949890137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.01,0.08299967765808106
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.01,0.08567872047424316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.01,0.09054656028747558
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,16,1,power_law_1.2,17.955601806640622
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.01,0.0932038402557373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.01,0.10545791625976561
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.01,0.11536447525024414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.01,0.09586048126220703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.01,0.138472318649292
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.01,0.16549055099487303
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.01,0.23077312469482422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.01,0.2814425659179688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.01,0.3947795104980469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.01,0.5084364700317383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,16,1,power_law_1.2,24.905805664062502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.01,0.7315193939208984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.01,0.031240320205688475
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.01,0.03329983949661255
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.01,0.0336844801902771
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.01,0.03492991924285889
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.01,0.04065792083740234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.01,0.04174784183502197
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.01,0.042271361351013184
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.01,0.04306496143341064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.01,0.04350207805633545
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.01,0.043924479484558104
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.01,0.04474815845489502
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.01,0.04558144092559814
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.01,0.0467903995513916
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.01,0.05092927932739257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.01,0.05420351982116699
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.01,0.05623807907104492
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.01,0.054558720588684086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.01,0.05840191841125488
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.01,0.9585196685791015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.01,0.06766016006469726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.01,0.09439807891845703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.01,0.08534527778625488
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.01,0.13624511718750001
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.01,0.16635583877563476
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.01,0.2372313690185547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,32,1,power_law_1.01,1.1832313537597656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.01,0.3121414375305176
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,16,1,power_law_1.2,40.318088378906246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.01,0.44984062194824215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.01,0.578328971862793
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.01,0.017745280265808107
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.01,0.019374719858169555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.01,0.019330559968948363
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.01,0.01977280020713806
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.01,0.019711999893188475
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.01,0.02004863977432251
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.01,0.019747840166091917
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.01,0.01997056007385254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.01,0.019782400131225585
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.01,0.01810367941856384
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,32,1,power_law_1.01,0.7203424072265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.01,0.018524160385131834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.01,0.02062079906463623
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.01,0.020631039142608644
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.01,0.022206079959869385
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.01,0.02283263921737671
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.01,0.024273281097412107
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.01,0.026899199485778808
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.01,0.0318016004562378
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.01,0.03774784088134765
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.01,0.04847743988037109
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.01,0.05643839836120605
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.01,0.07688191890716553
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.01,0.09472960472106934
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.01,0.12935232162475585
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.01,0.16717952728271485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,14336,2,8,32,1,power_law_1.2,0.0275545597076416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,14336,2,8,32,1,power_law_1.2,0.037447679042816165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,14336,2,8,32,1,power_law_1.2,0.04033408164978027
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,14336,2,8,32,1,power_law_1.2,0.03982144117355347
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,14336,2,8,32,1,power_law_1.2,0.06253376007080078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,14336,2,8,32,1,power_law_1.2,0.06339647769927978
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,14336,2,8,32,1,power_law_1.2,0.06518208026885987
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,14336,2,8,32,1,power_law_1.2,0.06638400077819824
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,14336,2,8,32,1,power_law_1.2,0.0652179193496704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,14336,2,8,32,1,power_law_1.2,0.07230144023895264
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,14336,2,8,32,1,power_law_1.2,0.07343359947204589
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,14336,2,8,32,1,power_law_1.2,0.08041664123535155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,14336,2,8,32,1,power_law_1.2,0.08282367706298828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,14336,2,8,32,1,power_law_1.2,0.08658495903015137
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,14336,2,8,32,1,power_law_1.2,0.09027263641357422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,14336,2,8,32,1,power_law_1.2,0.09470591545104981
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,14336,2,8,32,1,power_law_1.2,0.09561471939086914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,14336,2,8,32,1,power_law_1.2,0.1057260799407959
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,14336,2,8,32,1,power_law_1.2,0.11602432250976562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,14336,2,8,32,1,power_law_1.2,0.13889856338500978
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,14336,2,8,32,1,power_law_1.2,0.1657209587097168
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.01,0.24228992462158203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,14336,2,8,32,1,power_law_1.2,0.23113216400146483
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.01,0.31475391387939455
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,14336,2,8,32,1,power_law_1.2,0.28143552780151365
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,14336,2,8,32,1,power_law_1.2,0.3947520065307617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.01,0.39110145568847654
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,14336,2,8,32,1,power_law_1.2,0.5081740951538085
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,14336,2,8,32,1,power_law_1.2,0.030035200119018557
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,14336,2,8,32,1,power_law_1.2,0.03204607963562012
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,14336,2,8,32,1,power_law_1.2,0.03287807941436768
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,14336,2,8,32,1,power_law_1.2,0.034111359119415284
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,14336,2,8,32,1,power_law_1.2,0.03993151903152466
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,14336,2,8,32,1,power_law_1.2,0.04081024169921875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,14336,2,8,32,1,power_law_1.2,0.041057920455932616
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,14336,2,8,32,1,power_law_1.2,0.04190336227416992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,14336,2,8,32,1,power_law_1.2,0.042673921585083006
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,14336,2,8,32,1,power_law_1.2,0.04311103820800781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,14336,2,8,32,1,power_law_1.2,0.04397439956665039
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,14336,2,8,32,1,power_law_1.2,0.04476096153259278
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,14336,2,8,32,1,power_law_1.2,0.04600192070007324
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,14336,2,8,32,1,power_law_1.2,0.050542078018188476
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,14336,2,8,32,1,power_law_1.2,0.05428160190582275
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,14336,2,8,32,1,power_law_1.2,0.0566700792312622
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,14336,2,8,32,1,power_law_1.2,0.05445055961608887
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,14336,2,8,32,1,power_law_1.2,0.05895232200622559
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,14336,2,8,32,1,power_law_1.2,0.06937407970428466
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,14336,2,8,32,1,power_law_1.2,0.08211392402648926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,14336,2,8,32,1,power_law_1.2,0.09546239852905274
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,14336,2,8,32,1,power_law_1.2,0.7328915405273437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,14336,2,8,32,1,power_law_1.2,0.13621376037597657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,14336,2,8,32,1,power_law_1.2,0.16647615432739257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,14336,2,8,32,1,power_law_1.2,0.2384934425354004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,14336,2,8,32,1,power_law_1.2,0.9546451568603516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,14336,2,8,32,1,power_law_1.2,0.3122022438049316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,14336,2,8,32,1,power_law_1.2,1.1799827575683595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,4096,14336,2,8,32,1,power_law_1.2,0.017271679639816285
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,4096,14336,2,8,32,1,power_law_1.2,0.01807551980018616
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,4096,14336,2,8,32,1,power_law_1.2,0.0181222403049469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,4096,14336,2,8,32,1,power_law_1.2,0.019339519739151
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,4096,14336,2,8,32,1,power_law_1.2,0.019389439821243287
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,4096,14336,2,8,32,1,power_law_1.2,0.01976704001426697
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,4096,14336,2,8,32,1,power_law_1.2,0.01974143981933594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,4096,14336,2,8,32,1,power_law_1.2,0.019778560400009155
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,4096,14336,2,8,32,1,power_law_1.2,0.0197324800491333
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,4096,14336,2,8,32,1,power_law_1.2,0.019791359901428225
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,4096,14336,2,8,32,1,power_law_1.2,0.020166399478912352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,4096,14336,2,8,32,1,power_law_1.2,0.02054464101791382
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,4096,14336,2,8,32,1,power_law_1.2,0.02098560094833374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,4096,14336,2,8,32,1,power_law_1.2,0.02178623914718628
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,4096,14336,2,8,32,1,power_law_1.2,0.023795199394226075
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,4096,14336,2,8,32,1,power_law_1.2,0.02428031921386719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,14336,2,8,32,1,power_law_1.2,0.44932800292968744
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,4096,14336,2,8,32,1,power_law_1.2,0.027368319034576417
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,4096,14336,2,8,32,1,power_law_1.2,0.03243711948394775
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,14336,2,8,32,1,power_law_1.2,0.5782470321655273
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,4096,14336,2,8,32,1,power_law_1.2,0.05782464027404786
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,4096,14336,2,8,32,1,power_law_1.2,0.03797183990478516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,4096,14336,2,8,32,1,power_law_1.2,0.04849023818969726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,14336,2,8,32,1,power_law_1.2,0.7191340637207031
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,4096,14336,2,8,32,1,power_law_1.2,0.07827199935913086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,4096,14336,2,8,32,1,power_law_1.2,0.09840512275695801
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,4096,14336,2,8,32,1,power_law_1.2,0.13477375984191894
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,4096,14336,2,8,32,1,power_law_1.2,0.17418304443359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.01,0.049074559211730956
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.01,0.05992256164550781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.01,0.06149439811706543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.01,0.06895552158355714
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.01,0.10926015853881836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.01,0.11021120071411132
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.01,0.11128831863403321
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.01,0.11242239952087403
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.01,0.11304832458496095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.01,0.11454400062561035
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.01,0.11795519828796386
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.01,0.12009599685668945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.01,0.12390015602111817
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.01,0.12671104431152344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.01,0.12996992111206054
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.01,0.13426176071166993
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.01,0.13790080070495606
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.01,0.1523705577850342
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.01,0.16641536712646485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.01,0.19803647994995116
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.01,0.2581817626953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,4096,14336,2,8,32,1,power_law_1.2,0.25017791748046875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.01,0.35258689880371097
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.01,0.44035774230957025
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.01,0.6203488159179688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.01,0.7968224334716797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.01,1.1601837158203125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,4096,14336,2,8,32,1,power_law_1.2,0.3283635330200195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.01,0.03612735986709595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.01,0.0390393590927124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.01,0.03940799951553345
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.01,0.04273024082183838
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.01,0.05094016075134278
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.01,0.051686401367187496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.01,0.052197761535644524
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.01,1.5182591247558594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.01,0.05253695964813232
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.01,0.05311423778533936
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.01,0.05340735912322998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.01,0.05582655906677246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.01,0.05713471889495849
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.01,0.05832320213317871
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.01,0.06314367771148681
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.01,0.06487616062164306
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.01,0.06927680015563965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.01,0.0702790403366089
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.01,0.08120896339416503
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.01,0.09090496063232421
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.01,0.10890943527221679
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.01,0.12580927848815918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.01,0.18673728942871093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.01,0.22676223754882813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,4096,14336,2,8,32,1,power_law_1.2,0.40441921234130856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.01,0.3261804962158203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.01,0.42316734313964843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,32,1,power_law_1.01,1.8931500244140627
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.01,0.6147686386108399
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.01,0.7887609863281251
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.01,0.018585599660873413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.01,0.02220544099807739
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.01,0.022649600505828857
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.01,0.022648320198059083
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.01,0.02276927947998047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.01,0.02281791925430298
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.01,0.023036799430847167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.01,0.02363136053085327
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.01,0.024213120937347413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.01,0.019790079593658447
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,32,1,power_law_1.01,0.984674530029297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.01,0.019797120094299316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.01,0.024254720211029052
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.01,0.024639999866485594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.01,0.026241919994354247
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.01,0.028796160221099855
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.01,0.030377600193023678
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.01,0.03366015911102295
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.01,0.041125760078430176
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.01,0.05230783939361572
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.01,0.06453567981719971
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.01,0.07404543876647948
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.01,0.10025152206420898
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.01,0.1264038372039795
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.01,0.1799238395690918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.01,0.22947839736938475
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,16384,2,8,32,1,power_law_1.2,0.04694399833679199
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,16384,2,8,32,1,power_law_1.2,0.059024639129638665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,16384,2,8,32,1,power_law_1.2,0.05981696128845215
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,16384,2,8,32,1,power_law_1.2,0.06921664237976075
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,16384,2,8,32,1,power_law_1.2,0.10779647827148438
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,16384,2,8,32,1,power_law_1.2,0.10918975830078126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,16384,2,8,32,1,power_law_1.2,0.11055168151855468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,16384,2,8,32,1,power_law_1.2,0.11074111938476563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,16384,2,8,32,1,power_law_1.2,0.11218111991882325
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,16384,2,8,32,1,power_law_1.2,0.11324928283691407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,16384,2,8,32,1,power_law_1.2,0.11737471580505372
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,16384,2,8,32,1,power_law_1.2,0.1203878402709961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,16384,2,8,32,1,power_law_1.2,0.1224454402923584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,16384,2,8,32,1,power_law_1.2,0.1260582447052002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,16384,2,8,32,1,power_law_1.2,0.1295577621459961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,16384,2,8,32,1,power_law_1.2,0.13514368057250975
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,16384,2,8,32,1,power_law_1.2,0.1371564769744873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,16384,2,8,32,1,power_law_1.2,0.15163776397705078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,16384,2,8,32,1,power_law_1.2,0.16685951232910157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,16384,2,8,32,1,power_law_1.2,0.19873344421386718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.01,0.3357126235961914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,16384,2,8,32,1,power_law_1.2,0.2605619239807129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,16384,2,8,32,1,power_law_1.2,0.3530835342407227
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,16384,2,8,32,1,power_law_1.2,0.4421548843383789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.01,0.44094142913818357
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,16384,2,8,32,1,power_law_1.2,0.6190796661376953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,16384,2,8,32,1,power_law_1.2,0.7986681365966797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.01,0.5510675048828124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,16384,2,8,32,1,power_law_1.2,0.03619584083557129
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,16384,2,8,32,1,power_law_1.2,0.038620800971984864
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,16384,2,8,32,1,power_law_1.2,0.03898367881774902
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,16384,2,8,32,1,power_law_1.2,0.04423679828643799
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,16384,2,8,32,1,power_law_1.2,0.050912637710571286
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,16384,2,8,32,1,power_law_1.2,0.05171648025512695
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,16384,2,8,32,1,power_law_1.2,0.0525491189956665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,16384,2,8,32,1,power_law_1.2,0.05324672222137451
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,16384,2,8,32,1,power_law_1.2,0.053757438659667964
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,16384,2,8,32,1,power_law_1.2,0.054560642242431644
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,16384,2,8,32,1,power_law_1.2,0.05581759929656982
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,16384,2,8,32,1,power_law_1.2,0.05705152034759522
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,16384,2,8,32,1,power_law_1.2,0.0582047986984253
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,16384,2,8,32,1,power_law_1.2,0.06111936092376709
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,16384,2,8,32,1,power_law_1.2,0.06446335792541505
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,16384,2,8,32,1,power_law_1.2,0.06792831897735596
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,16384,2,8,32,1,power_law_1.2,0.07200255870819092
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,16384,2,8,32,1,power_law_1.2,0.08169471740722656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,16384,2,8,32,1,power_law_1.2,0.09237567901611328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,16384,2,8,32,1,power_law_1.2,0.10950847625732422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,16384,2,8,32,1,power_law_1.2,1.1580358123779297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,16384,2,8,32,1,power_law_1.2,0.12689279556274413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,16384,2,8,32,1,power_law_1.2,0.18696447372436525
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,16384,2,8,32,1,power_law_1.2,0.22799808502197266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,16384,2,8,32,1,power_law_1.2,1.5146739196777343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,16384,2,8,32,1,power_law_1.2,0.3262438583374023
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,16384,2,8,32,1,power_law_1.2,0.42580734252929686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,16384,2,8,32,1,power_law_1.2,1.8873516845703127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,6144,16384,2,8,32,1,power_law_1.2,0.018533120155334475
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,6144,16384,2,8,32,1,power_law_1.2,0.019326720237731933
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,6144,16384,2,8,32,1,power_law_1.2,0.02011712074279785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,6144,16384,2,8,32,1,power_law_1.2,0.02135551929473877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,6144,16384,2,8,32,1,power_law_1.2,0.021751039028167725
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,6144,16384,2,8,32,1,power_law_1.2,0.021756160259246825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,6144,16384,2,8,32,1,power_law_1.2,0.021817600727081297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,6144,16384,2,8,32,1,power_law_1.2,0.021784958839416505
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,6144,16384,2,8,32,1,power_law_1.2,0.021778559684753417
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,6144,16384,2,8,32,1,power_law_1.2,0.02243328094482422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,6144,16384,2,8,32,1,power_law_1.2,0.022225279808044434
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,6144,16384,2,8,32,1,power_law_1.2,0.024236159324645994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,6144,16384,2,8,32,1,power_law_1.2,0.024749441146850588
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,6144,16384,2,8,32,1,power_law_1.2,0.026737918853759767
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,16384,2,8,32,1,power_law_1.2,0.6151846313476563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,6144,16384,2,8,32,1,power_law_1.2,0.028822400569915772
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,6144,16384,2,8,32,1,power_law_1.2,0.0340883207321167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,16384,2,8,32,1,power_law_1.2,0.7893113708496093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,6144,16384,2,8,32,1,power_law_1.2,0.053197441101074217
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,6144,16384,2,8,32,1,power_law_1.2,0.03013887882232666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,6144,16384,2,8,32,1,power_law_1.2,0.04179840087890625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,6144,16384,2,8,32,1,power_law_1.2,0.06535744190216064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,6144,16384,2,8,32,1,power_law_1.2,0.07709760189056396
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,16384,2,8,32,1,power_law_1.2,0.9911885070800782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,6144,16384,2,8,32,1,power_law_1.2,0.10421952247619629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,6144,16384,2,8,32,1,power_law_1.2,0.13073023796081543
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,6144,16384,2,8,32,1,power_law_1.2,0.1875712013244629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,6144,16384,2,8,32,1,power_law_1.2,0.24018815994262693
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.01,0.041544318199157715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.01,0.045185279846191403
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.01,0.07671423912048339
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.01,0.09374591827392578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.01,0.11596416473388671
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.01,0.14813632011413574
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.01,0.16426944732666016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.01,0.171026554107666
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.01,0.17520959854125978
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.01,0.1790444755554199
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.01,0.18601728439331056
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.01,0.1991244888305664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.01,0.20669055938720704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.01,0.2268486404418945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.01,0.30145599365234377
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.01,0.30882495880126953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.01,0.3228070449829102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.01,0.3505561447143555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.01,0.37739646911621094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.01,0.4339936065673828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.01,0.4881887817382813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.01,0.6339884948730469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.01,0.7908185577392578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.01,1.038751983642578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.01,1.3166720581054687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.01,1.8377862548828126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,6144,16384,2,8,32,1,power_law_1.2,0.34918144226074216
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.01,2.3774278259277346
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.01,0.04390719890594482
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,32,1,power_law_1.01,2.923388671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.01,0.06487232208251953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.01,0.0938054370880127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.01,0.12984512329101563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.01,0.18083776473999022
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.01,0.05029695987701416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.01,0.21001087188720705
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.01,0.21843135833740232
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.01,0.22142847061157225
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.01,0.2254003143310547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.01,0.23300607681274413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.01,0.24199615478515626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.01,0.2447583961486816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.01,0.2525087928771973
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.01,0.29737600326538083
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.01,0.3036460876464844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.01,0.31485504150390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.01,0.3365024185180664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.01,0.36267967224121095
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.01,0.4118304061889648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.01,0.4766732788085937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.01,0.5814028930664062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.01,0.6809613037109374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.01,0.898062744140625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.01,1.1254528045654297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.01,1.5718112182617188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.01,1.9905421447753906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,32,1,power_law_1.01,2.439295349121094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,6144,16384,2,8,32,1,power_law_1.2,0.4601299285888672
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.01,0.07097216129302979
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.01,0.09707712173461915
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.01,0.07390272140502929
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.01,0.11342528343200683
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.01,0.12789440155029297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.01,0.15756287574768066
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.01,0.1643244743347168
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.01,0.1667123222351074
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.01,0.1694246482849121
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.01,0.17700735092163086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.01,0.1833299255371094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.01,0.1874220848083496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.01,0.19492095947265625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.01,0.2037260818481445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.01,0.21078464508056644
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.01,0.08386240005493165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.01,0.22674560546875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,6144,16384,2,8,32,1,power_law_1.2,0.5695987319946288
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.01,0.26775360107421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.01,0.30297983169555665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.01,0.3852339172363281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.01,0.46454910278320316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.01,0.6406034851074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.01,0.8204192352294921
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.01,1.1901516723632812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.01,1.5846694946289062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,256,32,1,power_law_1.2,0.04185408115386963
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,256,32,1,power_law_1.2,0.04476223945617676
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,256,32,1,power_law_1.2,0.07654335975646973
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,256,32,1,power_law_1.2,0.0904032039642334
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,256,32,1,power_law_1.2,0.11173055648803712
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,256,32,1,power_law_1.2,0.1418227195739746
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,256,32,1,power_law_1.2,0.16328895568847657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,256,32,1,power_law_1.2,0.16825279235839843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,256,32,1,power_law_1.2,0.17272895812988281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,256,32,1,power_law_1.2,0.17884159088134766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,256,32,1,power_law_1.2,0.18666175842285154
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,256,32,1,power_law_1.2,0.19783296585083007
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,256,32,1,power_law_1.2,0.206376953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,256,32,1,power_law_1.2,0.23002559661865235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,256,32,1,power_law_1.2,0.30276672363281254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,256,32,1,power_law_1.2,0.3113715171813965
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,256,32,1,power_law_1.2,0.3246803283691406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,256,32,1,power_law_1.2,0.353067512512207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,256,32,1,power_law_1.2,0.3806284713745117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,256,32,1,power_law_1.2,0.43676609039306646
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,256,32,1,power_law_1.2,0.49032062530517584
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,256,32,1,power_law_1.2,0.6382944107055664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,256,32,1,power_law_1.2,0.7940332794189453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,256,32,1,power_law_1.2,1.041137924194336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,12288,7168,2048,8,256,32,1,power_law_1.01,2.3621900939941405
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,256,32,1,power_law_1.2,1.3193011474609375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16384,7168,2048,8,256,32,1,power_law_1.01,3.0445959472656248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,20480,7168,2048,8,256,32,1,power_law_1.01,3.8682220458984373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,256,32,1,power_law_1.2,1.8477043151855468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,256,32,1,power_law_1.2,0.05002751827239991
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,256,32,1,power_law_1.2,2.3900825500488283
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,256,32,1,power_law_1.2,0.06908351898193359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,256,32,1,power_law_1.2,0.04311232089996338
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,256,32,1,power_law_1.2,0.09127743721008301
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,256,32,1,power_law_1.2,0.12612735748291015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,256,32,1,power_law_1.2,0.17572799682617188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,256,32,1,power_law_1.2,0.20902271270751954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,256,32,1,power_law_1.2,2.9469445800781253
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,256,32,1,power_law_1.2,0.21570688247680664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,256,32,1,power_law_1.2,0.21892480850219725
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,256,32,1,power_law_1.2,0.22609535217285157
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,256,32,1,power_law_1.2,0.23088960647583007
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,256,32,1,power_law_1.2,0.23958335876464845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,256,32,1,power_law_1.2,0.2447372817993164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,256,32,1,power_law_1.2,0.2548716735839844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,256,32,1,power_law_1.2,0.30015743255615235
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,256,32,1,power_law_1.2,0.30976703643798825
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,256,32,1,power_law_1.2,0.3201702499389648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,256,32,1,power_law_1.2,0.34860801696777344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,256,32,1,power_law_1.2,0.3734707260131836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,256,32,1,power_law_1.2,0.43138111114501954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,256,32,1,power_law_1.2,0.4932211303710938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,256,32,1,power_law_1.2,0.5967116928100585
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,256,32,1,power_law_1.2,0.7178304290771484
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,256,32,1,power_law_1.2,0.9116108703613282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,256,32,1,power_law_1.2,1.1402105712890624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,256,32,1,power_law_1.2,1.5808531188964845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,256,32,1,power_law_1.2,0.07064832210540771
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,256,32,1,power_law_1.2,2.023638458251953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,256,32,1,power_law_1.2,0.08296640396118164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,256,32,1,power_law_1.2,0.09468480110168456
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,256,32,1,power_law_1.2,2.4550489807128906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,256,32,1,power_law_1.2,0.10927167892456055
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,256,32,1,power_law_1.2,0.12420928001403808
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,256,32,1,power_law_1.2,0.15644672393798828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,256,32,1,power_law_1.2,0.16266048431396485
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,256,32,1,power_law_1.2,0.16616064071655273
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,256,32,1,power_law_1.2,0.17233407974243165
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,256,32,1,power_law_1.2,0.17754943847656252
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,256,32,1,power_law_1.2,0.18491647720336915
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,256,32,1,power_law_1.2,0.1915827178955078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,256,32,1,power_law_1.2,0.20332992553710935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,256,32,1,power_law_1.2,0.21350271224975587
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,256,32,1,power_law_1.2,0.22694143295288086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,256,32,1,power_law_1.2,0.07387712001800537
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,256,32,1,power_law_1.2,0.2526028823852539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,256,32,1,power_law_1.2,0.30928192138671873
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,256,32,1,power_law_1.2,0.372476806640625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,256,32,1,power_law_1.2,0.4971750259399414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,256,32,1,power_law_1.2,0.6175289535522461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,256,32,1,power_law_1.2,0.9382611083984376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,256,32,1,power_law_1.2,1.2379161834716796
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,256,32,1,power_law_1.2,1.8404826354980468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.01,0.034897921085357664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.01,0.051698560714721675
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.01,0.057793917655944826
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.01,0.0656108808517456
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.01,0.07846784114837646
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.01,0.07838592052459717
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.01,0.08037247657775878
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.01,0.08387264251708984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.01,0.08543999671936035
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.01,0.09098943710327148
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.01,0.12811264038085937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.01,0.13086655616760254
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.01,0.13584896087646486
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.01,0.13835391998291016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.01,0.14249088287353515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.01,0.15114687919616698
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.01,0.16907072067260742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8192,7168,2048,8,256,32,1,power_law_1.2,2.516809539794922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.01,0.18414848327636718
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.01,0.2530348777770996
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.01,0.21770303726196288
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.01,0.3287654495239258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.01,0.41338943481445317
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.01,0.5712070465087891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.01,0.7286854553222656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.01,1.0435391998291015
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.01,1.3617190551757812
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.01,0.03496000051498413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.01,0.04020671844482422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.01,0.047263360023498534
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.01,0.05550335884094239
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,32,1,power_law_1.01,1.6796453857421874
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.01,0.0703334379196167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.01,0.08554112434387207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.01,0.08554176330566407
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.01,0.0878700828552246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.01,0.08240384101867676
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.01,0.08708160400390626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.01,0.09038847923278809
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.01,0.10605567932128905
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.01,0.10862655639648437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.01,0.11393600463867189
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.01,0.11674367904663085
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.01,0.11889535903930663
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.01,0.12906815528869628
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.01,0.14117055892944336
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.01,0.16681535720825197
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.01,0.19959039688110353
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.01,0.23538879394531248
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.01,0.2975116729736328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.01,0.3725273513793945
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.01,0.5116019058227539
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.01,0.6360575866699218
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.01,0.9094207763671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.01,1.184456329345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,32,1,power_law_1.01,1.4490098571777343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,1536,8,128,32,1,power_law_1.2,0.030424320697784425
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,1536,8,128,32,1,power_law_1.2,0.05073408126831055
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,1536,8,128,32,1,power_law_1.2,0.03514944076538086
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,1536,8,128,32,1,power_law_1.2,0.05620992183685303
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,1536,8,128,32,1,power_law_1.2,0.06469823837280272
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,1536,8,128,32,1,power_law_1.2,0.07888768196105958
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,1536,8,128,32,1,power_law_1.2,0.08002112388610841
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,1536,8,128,32,1,power_law_1.2,0.08141887664794922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,1536,8,128,32,1,power_law_1.2,0.08498751640319824
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,1536,8,128,32,1,power_law_1.2,0.09055104255676269
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,1536,8,128,32,1,power_law_1.2,0.12825920104980468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,1536,8,128,32,1,power_law_1.2,0.1368838405609131
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,1536,8,128,32,1,power_law_1.2,0.13943872451782227
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,1536,8,128,32,1,power_law_1.2,0.14297984123229981
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,1536,8,128,32,1,power_law_1.2,0.15187199592590334
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,1536,8,128,32,1,power_law_1.2,0.16972415924072265
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,1536,8,128,32,1,power_law_1.2,0.18543424606323242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,1536,8,128,32,1,power_law_1.2,0.08248767852783204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,1536,8,128,32,1,power_law_1.2,0.2185068893432617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,1536,8,128,32,1,power_law_1.2,0.127640962600708
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,1536,8,128,32,1,power_law_1.2,0.25443712234497073
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,1536,8,128,32,1,power_law_1.2,0.32651329040527344
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,1536,8,128,32,1,power_law_1.2,0.4184249496459961
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,1536,8,128,32,1,power_law_1.2,0.5704044723510743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,1536,8,128,32,1,power_law_1.2,0.7285791778564453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,1536,8,128,32,1,power_law_1.2,0.03500927925109863
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,1536,8,128,32,1,power_law_1.2,0.04023551940917969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,1536,8,128,32,1,power_law_1.2,0.0467244815826416
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,1536,8,128,32,1,power_law_1.2,0.05407999992370606
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,1536,8,128,32,1,power_law_1.2,0.07026368141174316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,1536,8,128,32,1,power_law_1.2,0.08511615753173828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,1536,8,128,32,1,power_law_1.2,0.08438719749450684
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,1536,8,128,32,1,power_law_1.2,0.08563967704772948
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,1536,8,128,32,1,power_law_1.2,0.08782719612121583
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,1536,8,128,32,1,power_law_1.2,0.08937088012695313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,1536,8,128,32,1,power_law_1.2,0.09109248161315918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,1536,8,128,32,1,power_law_1.2,0.10547519683837892
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,1536,8,128,32,1,power_law_1.2,0.10974271774291992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,1536,8,128,32,1,power_law_1.2,0.11591744422912598
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,1536,8,128,32,1,power_law_1.2,0.11851967811584471
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,1536,8,128,32,1,power_law_1.2,0.12257023811340333
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,1536,8,128,32,1,power_law_1.2,0.13041983604431154
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,1536,8,128,32,1,power_law_1.2,0.14616512298583983
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,1536,8,128,32,1,power_law_1.2,0.17208255767822264
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,1536,8,128,32,1,power_law_1.2,0.20560384750366212
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,1536,8,128,32,1,power_law_1.2,0.23737728118896487
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,1536,8,128,32,1,power_law_1.2,0.2999807929992676
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,1536,8,128,32,1,power_law_1.2,0.37673343658447267
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,1536,8,128,32,1,power_law_1.2,1.0446131134033203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,1536,8,128,32,1,power_law_1.2,0.5134534454345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,1536,8,128,32,1,power_law_1.2,0.6383590316772461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,1536,8,128,32,1,power_law_1.2,1.3647257995605468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,1536,8,128,32,1,power_law_1.2,0.9088940429687501
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,1536,8,128,32,1,power_law_1.2,1.6806661987304685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,1536,8,128,32,1,power_law_1.2,1.1821279907226563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,1536,8,128,32,1,power_law_1.2,1.4495237731933595
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.01,0.040282239913940435
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.01,0.05597631931304932
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.01,0.07541247844696045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.01,0.08981311798095704
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.01,0.10996735572814942
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.01,0.14145983695983885
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.01,0.13521727561950683
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.01,0.1402783966064453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.01,0.14398591995239257
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.01,0.14837504386901856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.01,0.15911935806274413
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.01,0.16854591369628907
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.01,0.2128473663330078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.01,0.21999040603637696
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.01,0.2234560012817383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.01,0.22927360534667968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.01,0.24397312164306642
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.01,0.2699993515014648
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.01,0.2967193603515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.01,0.35296001434326174
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.01,0.4392537689208984
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.01,0.5850758361816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.01,0.7098457336425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.01,0.9767648315429687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.01,1.2479564666748046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.01,1.785529022216797
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.01,2.3288294982910154
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.01,0.043576960563659665
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,32,1,power_law_1.01,2.873457336425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.01,0.068504319190979
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.01,0.08511551856994629
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.01,0.11308927536010742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.01,0.1542956829071045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.01,0.14766528129577636
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.01,0.15222271919250488
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.01,0.15574591636657714
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.01,0.1581753635406494
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.01,0.1619718360900879
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.01,0.1734822463989258
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.01,0.048659839630126954
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.01,0.19084991455078124
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.01,0.2042239952087402
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.01,0.20505279541015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.01,0.1954207992553711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.01,0.22715135574340822
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.01,0.2488185691833496
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.01,0.26985471725463867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.01,0.31436031341552734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.01,0.36769214630126956
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.01,0.4736320114135742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.01,0.573765754699707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.01,0.785528335571289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.01,0.9890035247802734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.01,1.3836901855468748
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.01,1.8055290222167968
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,32,1,power_law_1.01,2.2266636657714844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,6144,2560,8,160,32,1,power_law_1.2,0.04433023929595947
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,6144,2560,8,160,32,1,power_law_1.2,0.06578239917755127
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,6144,2560,8,160,32,1,power_law_1.2,0.07459839820861816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,6144,2560,8,160,32,1,power_law_1.2,0.08639039993286132
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,6144,2560,8,160,32,1,power_law_1.2,0.102359037399292
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,6144,2560,8,160,32,1,power_law_1.2,0.13974080085754395
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,6144,2560,8,160,32,1,power_law_1.2,0.13548095703125002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,6144,2560,8,160,32,1,power_law_1.2,0.1405068778991699
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,6144,2560,8,160,32,1,power_law_1.2,0.1462611198425293
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,6144,2560,8,160,32,1,power_law_1.2,0.15093119621276857
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,6144,2560,8,160,32,1,power_law_1.2,0.16104703903198242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,6144,2560,8,160,32,1,power_law_1.2,0.1700396728515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,6144,2560,8,160,32,1,power_law_1.2,0.2141152000427246
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,6144,2560,8,160,32,1,power_law_1.2,0.22044351577758786
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,6144,2560,8,160,32,1,power_law_1.2,0.22640575408935545
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,6144,2560,8,160,32,1,power_law_1.2,0.23082752227783204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,6144,2560,8,160,32,1,power_law_1.2,0.24510143280029295
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,6144,2560,8,160,32,1,power_law_1.2,0.27054208755493164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,6144,2560,8,160,32,1,power_law_1.2,0.29817920684814453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,6144,2560,8,160,32,1,power_law_1.2,0.36398849487304685
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,6144,2560,8,160,32,1,power_law_1.2,0.4399302291870117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,6144,2560,8,160,32,1,power_law_1.2,0.5865881729125977
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,6144,2560,8,160,32,1,power_law_1.2,0.7134694671630859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,6144,2560,8,160,32,1,power_law_1.2,0.9796281433105468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,6144,2560,8,160,32,1,power_law_1.2,1.2504108428955079
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,6144,2560,8,160,32,1,power_law_1.2,0.044803838729858395
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,6144,2560,8,160,32,1,power_law_1.2,0.04974080085754394
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,6144,2560,8,160,32,1,power_law_1.2,0.0672102403640747
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,6144,2560,8,160,32,1,power_law_1.2,0.08152383804321289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,6144,2560,8,160,32,1,power_law_1.2,0.10876031875610352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,6144,2560,8,160,32,1,power_law_1.2,0.15286272048950195
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,6144,2560,8,160,32,1,power_law_1.2,0.14699456214904785
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,6144,2560,8,160,32,1,power_law_1.2,0.15237183570861818
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,6144,2560,8,160,32,1,power_law_1.2,0.15901439666748046
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,6144,2560,8,160,32,1,power_law_1.2,0.15930944442749023
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,6144,2560,8,160,32,1,power_law_1.2,0.1662156867980957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,6144,2560,8,160,32,1,power_law_1.2,0.1705638313293457
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,6144,2560,8,160,32,1,power_law_1.2,0.20546176910400388
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,6144,2560,8,160,32,1,power_law_1.2,0.20579904556274414
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,6144,2560,8,160,32,1,power_law_1.2,0.21581632614135743
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,6144,2560,8,160,32,1,power_law_1.2,1.78625732421875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,6144,2560,8,160,32,1,power_law_1.2,0.22200576782226564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,6144,2560,8,160,32,1,power_law_1.2,0.230446720123291
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,6144,2560,8,160,32,1,power_law_1.2,0.2513830375671387
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,6144,2560,8,160,32,1,power_law_1.2,0.2709555244445801
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,6144,2560,8,160,32,1,power_law_1.2,0.3222758483886719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,6144,2560,8,160,32,1,power_law_1.2,0.3723129653930664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,6144,2560,8,160,32,1,power_law_1.2,0.4880422210693359
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,6144,2560,8,160,32,1,power_law_1.2,0.5792780685424804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,6144,2560,8,160,32,1,power_law_1.2,0.7900627136230469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,6144,2560,8,160,32,1,power_law_1.2,0.9926335906982422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,6144,2560,8,160,32,1,power_law_1.2,2.331853485107422
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,6144,2560,8,160,32,1,power_law_1.2,1.3894143676757813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,6144,2560,8,160,32,1,power_law_1.2,2.881628112792969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,6144,2560,8,160,32,1,power_law_1.2,1.8124114990234375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,6144,2560,8,160,32,1,power_law_1.2,2.225015106201172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,7168,2048,8,384,32,1,power_law_1.01,0.0460422420501709
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.01,0.05012928009033203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.01,0.06806079864501953
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.01,0.09712512016296386
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.01,0.12924863815307616
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.01,0.17126592636108398
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.01,0.1849555206298828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.01,0.21293312072753906
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.01,0.22171903610229493
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.01,0.23140544891357423
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.01,0.23812992095947266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.01,0.24362176895141602
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.01,0.25821504592895506
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.01,0.2811884880065918
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.01,0.2986835289001465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.01,0.3171628761291504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.01,0.4040243148803711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.01,0.4337811279296875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.01,0.46156032562255855
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.01,0.5922137451171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.01,0.7781817626953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.01,0.8943622589111329
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.01,1.1689810943603516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.01,1.4461106872558593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.01,2.0080979919433597
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.01,0.047294721603393555
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,32,1,power_law_1.01,3.135704345703125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.01,0.05252287864685059
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.01,0.09814656257629395
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.01,0.06398784160614014
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.01,0.14542207717895508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.01,0.21474048614501956
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.01,0.24815935134887696
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.01,0.28449344635009766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.01,0.2942086410522461
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.01,0.30230207443237306
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.01,0.31733568191528316
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.01,0.31286399841308593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.01,0.329736328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.01,0.3481017684936524
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.01,0.3461030578613281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.01,0.3619865417480469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.01,0.42879680633544925
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.01,0.44986049652099613
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.01,0.4742534255981445
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.01,0.5280217742919922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.01,0.5808505630493164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.01,0.696266860961914
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.01,0.8147456359863281
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.01,1.0519232177734374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.01,1.2736121368408202
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.01,1.7311827087402343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.01,2.1927871704101562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,32,1,power_law_1.01,2.6636468505859376
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.01,0.08939264297485353
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.01,0.09057855606079102
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.01,0.08930879592895508
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.01,0.09122112274169922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.01,0.09425855636596679
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.01,0.1004601573944092
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.01,0.10705408096313476
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.01,0.11278207778930664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.01,0.1192307186126709
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.01,0.12538368225097657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.01,0.13735679626464842
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.01,0.15682175636291504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.01,0.17061119079589843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.01,0.2089593505859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.01,0.272857608795166
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.01,0.3188678359985352
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.01,0.4028160095214844
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.01,0.5623231887817383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.01,0.7168428802490234
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.01,1.0368883514404297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.01,1.3477183532714843
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.01,1.9972364807128904
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.01,2.6174456787109373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.01,3.9180505371093752
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,7168,2048,8,384,32,1,power_law_1.2,0.04766272068023682
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,7168,2048,8,384,32,1,power_law_1.2,0.0794163179397583
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,7168,2048,8,384,32,1,power_law_1.2,0.09660863876342773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,7168,2048,8,384,32,1,power_law_1.2,0.123373441696167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,7168,2048,8,384,32,1,power_law_1.2,0.15721280097961426
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,7168,2048,8,384,32,1,power_law_1.2,0.17716032028198242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,7168,2048,8,384,32,1,power_law_1.2,0.2126374435424805
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,7168,2048,8,384,32,1,power_law_1.2,0.2187910461425781
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,7168,2048,8,384,32,1,power_law_1.2,0.22489471435546876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,7168,2048,8,384,32,1,power_law_1.2,0.232992000579834
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,7168,2048,8,384,32,1,power_law_1.2,0.24862016677856444
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,7168,2048,8,384,32,1,power_law_1.2,0.2578016090393066
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,7168,2048,8,384,32,1,power_law_1.2,0.28315967559814453
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,7168,2048,8,384,32,1,power_law_1.2,0.30444032669067383
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,7168,2048,8,384,32,1,power_law_1.2,0.31820608139038087
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,7168,2048,8,384,32,1,power_law_1.2,0.40571968078613285
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,7168,2048,8,384,32,1,power_law_1.2,0.43613822937011715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,7168,2048,8,384,32,1,power_law_1.2,0.46415550231933594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,7168,2048,8,384,32,1,power_law_1.2,0.5302086257934571
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,7168,2048,8,384,32,1,power_law_1.2,0.6059980773925782
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,7168,2048,8,384,32,1,power_law_1.2,0.7809747314453125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,7168,2048,8,384,32,1,power_law_1.2,0.8974797058105469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,7168,2048,8,384,32,1,power_law_1.2,1.1842655944824219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,7168,2048,8,384,32,1,power_law_1.2,1.448901824951172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,7168,2048,8,384,32,1,power_law_1.2,2.0126130676269534
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,7168,2048,8,384,32,1,power_law_1.2,2.580602111816406
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,7168,2048,8,384,32,1,power_law_1.2,3.153367004394531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,7168,2048,8,384,32,1,power_law_1.2,0.052529277801513674
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,7168,2048,8,384,32,1,power_law_1.2,0.06869887828826904
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,7168,2048,8,384,32,1,power_law_1.2,0.09833087921142578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,7168,2048,8,384,32,1,power_law_1.2,0.13885503768920898
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,7168,2048,8,384,32,1,power_law_1.2,0.19889471054077149
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,7168,2048,8,384,32,1,power_law_1.2,0.23775167465209962
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,7168,2048,8,384,32,1,power_law_1.2,0.2814713668823242
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,7168,2048,8,384,32,1,power_law_1.2,0.2897638320922852
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,7168,2048,8,384,32,1,power_law_1.2,0.044156799316406245
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,7168,2048,8,384,32,1,power_law_1.2,0.29623039245605465
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,7168,2048,8,384,32,1,power_law_1.2,0.300385913848877
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,7168,2048,8,384,32,1,power_law_1.2,0.31619712829589847
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,7168,2048,8,384,32,1,power_law_1.2,0.3218521499633789
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,7168,2048,8,384,32,1,power_law_1.2,0.3432704162597656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,7168,2048,8,384,32,1,power_law_1.2,0.3483795166015625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,7168,2048,8,384,32,1,power_law_1.2,0.3621152114868164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,7168,2048,8,384,32,1,power_law_1.2,0.4264640045166016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,7168,2048,8,384,32,1,power_law_1.2,0.45884735107421876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,7168,2048,8,384,32,1,power_law_1.2,0.48490432739257816
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,7168,2048,8,384,32,1,power_law_1.2,0.5420710372924804
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,7168,2048,8,384,32,1,power_law_1.2,0.6002579116821289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,7168,2048,8,384,32,1,power_law_1.2,0.7335794830322266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,7168,2048,8,384,32,1,power_law_1.2,0.8406739044189454
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,7168,2048,8,384,32,1,power_law_1.2,1.0741299438476564
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,7168,2048,8,384,32,1,power_law_1.2,1.3066943359374998
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,7168,2048,8,384,32,1,power_law_1.2,1.7627655029296876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,7168,2048,8,384,32,1,power_law_1.2,2.223222351074219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2,7168,2048,8,384,32,1,power_law_1.2,0.08950912475585937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,8,7168,2048,8,384,32,1,power_law_1.2,0.09140224456787109
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,16,7168,2048,8,384,32,1,power_law_1.2,0.09426048278808594
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,7168,2048,8,384,32,1,power_law_1.2,2.6918795776367186
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,32,7168,2048,8,384,32,1,power_law_1.2,0.100447359085083
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,64,7168,2048,8,384,32,1,power_law_1.2,0.11245951652526856
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,80,7168,2048,8,384,32,1,power_law_1.2,0.11923328399658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,96,7168,2048,8,384,32,1,power_law_1.2,0.12538304328918456
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,128,7168,2048,8,384,32,1,power_law_1.2,0.1374611186981201
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,160,7168,2048,8,384,32,1,power_law_1.2,0.15498496055603028
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,192,7168,2048,8,384,32,1,power_law_1.2,0.17271360397338867
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,256,7168,2048,8,384,32,1,power_law_1.2,0.2096467208862305
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1,7168,2048,8,384,32,1,power_law_1.2,0.0884921646118164
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,320,7168,2048,8,384,32,1,power_law_1.2,0.27902719497680667
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4,7168,2048,8,384,32,1,power_law_1.2,0.09019904136657715
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,48,7168,2048,8,384,32,1,power_law_1.2,0.1061081600189209
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,384,7168,2048,8,384,32,1,power_law_1.2,0.313317756652832
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,768,7168,2048,8,384,32,1,power_law_1.2,0.5610156631469726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,512,7168,2048,8,384,32,1,power_law_1.2,0.4034534454345703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1024,7168,2048,8,384,32,1,power_law_1.2,0.7210253143310548
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,1536,7168,2048,8,384,32,1,power_law_1.2,1.0321785736083986
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,2048,7168,2048,8,384,32,1,power_law_1.2,1.3484480285644531
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,3072,7168,2048,8,384,32,1,power_law_1.2,1.9910418701171875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,4096,7168,2048,8,384,32,1,power_law_1.2,2.6382733154296876
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.01,0.02923327922821045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.01,0.0341209602355957
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.01,0.0406931209564209
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.01,0.05898303985595703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.01,0.06854207992553711
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.01,0.08953472137451171
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.01,0.09513664245605469
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.01,0.09943039894104004
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.01,0.10367808341979981
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.01,0.107260160446167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.01,0.11426176071166991
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.01,0.1256486415863037
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.01,0.12807807922363282
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.01,0.1325376033782959
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.01,0.13316415786743163
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.01,0.13773247718811035
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.01,0.1455014419555664
SGLang,0.5.9,NVIDIA GB300,moe,sglang_flashinfer_cutedsl_moe,nvfp4,6144,7168,2048,8,384,32,1,power_law_1.2,3.9049600219726566
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.01,0.17836223602294923
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.01,0.2292582321166992
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.01,0.27061824798583983
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.01,0.3303699111938477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.01,0.4180704116821289
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.01,0.16154495239257813
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.01,0.5491648101806641
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.01,0.7003372955322266
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.01,1.0110336303710938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.01,1.315087432861328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,32,1,power_law_1.01,1.6096780395507815
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,2688,1856,6,128,32,1,power_law_1.2,0.033775360584259034
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,2688,1856,6,128,32,1,power_law_1.2,0.04537024021148682
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,2688,1856,6,128,32,1,power_law_1.2,0.055557761192321774
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,2688,1856,6,128,32,1,power_law_1.2,0.06709760189056396
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,2688,1856,6,128,32,1,power_law_1.2,0.0900166416168213
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,2688,1856,6,128,32,1,power_law_1.2,0.09662464141845703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,2688,1856,6,128,32,1,power_law_1.2,0.10177599906921386
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,2688,1856,6,128,32,1,power_law_1.2,0.02834752082824707
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,2688,1856,6,128,32,1,power_law_1.2,0.10525119781494141
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,2688,1856,6,128,32,1,power_law_1.2,0.11696255683898926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,2688,1856,6,128,32,1,power_law_1.2,0.12791680335998534
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,2688,1856,6,128,32,1,power_law_1.2,0.13021504402160644
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,2688,1856,6,128,32,1,power_law_1.2,0.13485568046569824
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,2688,1856,6,128,32,1,power_law_1.2,0.1360147190093994
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,2688,1856,6,128,32,1,power_law_1.2,0.10924223899841308
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,2688,1856,6,128,32,1,power_law_1.2,0.1402195167541504
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,2688,1856,6,128,32,1,power_law_1.2,0.14733823776245117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,2688,1856,6,128,32,1,power_law_1.2,0.17904319763183593
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,2688,1856,6,128,32,1,power_law_1.2,0.16447168350219726
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,2688,1856,6,128,32,1,power_law_1.2,0.23797952651977541
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,2688,1856,6,128,32,1,power_law_1.2,0.2714515113830566
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,2688,1856,6,128,32,1,power_law_1.2,0.33264190673828126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,2688,1856,6,128,32,1,power_law_1.2,0.4170054244995117
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,2688,1856,6,128,32,1,power_law_1.2,0.5484140777587891
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,2688,1856,6,128,32,1,power_law_1.2,0.7041747283935547
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,2688,1856,6,128,32,1,power_law_1.2,1.0115174102783204
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,2688,1856,6,128,32,1,power_law_1.2,1.3257452392578126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.01,0.05828800201416016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.01,0.09363583564758302
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.01,0.14543423652648926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.01,0.22199552536010742
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.01,0.31014783859252926
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.01,0.5517900848388673
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,2688,1856,6,128,32,1,power_law_1.2,1.6139634704589845
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.01,0.573936653137207
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.01,0.6345363235473632
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.01,0.5907392120361328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.01,0.6814035034179687
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.01,0.7265721893310546
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.01,0.5292383956909179
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.01,0.8384146881103515
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.01,0.9001407623291016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.01,0.9982675170898437
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.01,1.1913107299804688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.01,0.9360134124755859
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.01,1.0256050872802736
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.01,0.49556545257568363
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.01,1.2684799957275392
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.01,1.5145458984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.01,2.018746185302734
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.01,3.472879333496094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.01,2.47835205078125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.01,4.436581115722656
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.01,6.407550659179688
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.01,8.3892626953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,32,1,power_law_1.01,10.382171630859375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.01,0.5152576065063477
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.01,0.26387968063354494
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.01,0.7061612701416016
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.01,1.64385986328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.01,0.1473523235321045
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.01,1.8070700073242185
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.01,1.6138156127929686
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.01,1.8804045104980467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.01,1.8108486938476562
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.01,1.9781312561035158
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.01,1.8918202209472657
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.01,2.079594879150391
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.01,2.017229461669922
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.01,2.260042266845703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.01,2.28770751953125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.01,0.9628768157958983
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.01,4.016694946289062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.01,1.7623289489746092
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.01,3.219584655761719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.01,4.904854431152343
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.01,5.7466888427734375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.01,6.221920776367187
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.01,11.314434814453126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.01,14.427349853515626
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.01,15.299534912109374
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1,4096,2688,22,512,32,1,power_law_1.2,0.06039103984832763
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2,4096,2688,22,512,32,1,power_law_1.2,0.09733951568603516
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4,4096,2688,22,512,32,1,power_law_1.2,0.14653951644897462
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8,4096,2688,22,512,32,1,power_law_1.2,0.20912832260131836
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.01,7.166807250976563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16,4096,2688,22,512,32,1,power_law_1.2,0.284323844909668
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,48,4096,2688,22,512,32,1,power_law_1.2,0.5294259262084962
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,32,4096,2688,22,512,32,1,power_law_1.2,0.48594558715820313
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,64,4096,2688,22,512,32,1,power_law_1.2,0.5462758255004883
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,80,4096,2688,22,512,32,1,power_law_1.2,0.5780492782592773
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,96,4096,2688,22,512,32,1,power_law_1.2,0.6128364944458008
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,128,4096,2688,22,512,32,1,power_law_1.2,0.648629150390625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,160,4096,2688,22,512,32,1,power_law_1.2,0.7064793395996094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,192,4096,2688,22,512,32,1,power_law_1.2,0.7528934478759766
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,256,4096,2688,22,512,32,1,power_law_1.2,0.862005157470703
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.01,25.57579833984375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,320,4096,2688,22,512,32,1,power_law_1.2,0.936828155517578
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,384,4096,2688,22,512,32,1,power_law_1.2,1.029710693359375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,768,4096,2688,22,512,32,1,power_law_1.2,0.9594860839843751
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,512,4096,2688,22,512,32,1,power_law_1.2,1.2194355010986329
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1024,4096,2688,22,512,32,1,power_law_1.2,1.0530278778076172
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,1536,4096,2688,22,512,32,1,power_law_1.2,1.3116365051269532
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,2048,4096,2688,22,512,32,1,power_law_1.2,1.5609164428710938
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,3072,4096,2688,22,512,32,1,power_law_1.2,2.0351417541503904
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,4096,4096,2688,22,512,32,1,power_law_1.2,2.5034713745117188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,6144,4096,2688,22,512,32,1,power_law_1.2,3.498239440917969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,32,1,power_law_1.01,31.9095361328125
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,8192,4096,2688,22,512,32,1,power_law_1.2,4.4550912475585935
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,12288,4096,2688,22,512,32,1,power_law_1.2,6.459616088867188
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2,4096,2688,22,512,32,1,power_law_1.2,0.24658367156982425
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4,4096,2688,22,512,32,1,power_law_1.2,0.4368467330932617
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1,4096,2688,22,512,32,1,power_law_1.2,0.142807035446167
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16,4096,2688,22,512,32,1,power_law_1.2,0.8942918395996093
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,32,4096,2688,22,512,32,1,power_law_1.2,1.630438995361328
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,48,4096,2688,22,512,32,1,power_law_1.2,1.7589273071289064
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,64,4096,2688,22,512,32,1,power_law_1.2,1.8345011901855468
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,80,4096,2688,22,512,32,1,power_law_1.2,1.900437774658203
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,96,4096,2688,22,512,32,1,power_law_1.2,1.9412882995605467
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,16384,4096,2688,22,512,32,1,power_law_1.2,8.453369750976563
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8,4096,2688,22,512,32,1,power_law_1.2,0.6274272155761719
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,128,4096,2688,22,512,32,1,power_law_1.2,1.9624147033691408
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,192,4096,2688,22,512,32,1,power_law_1.2,2.031676788330078
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,256,4096,2688,22,512,32,1,power_law_1.2,2.106361541748047
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,320,4096,2688,22,512,32,1,power_law_1.2,2.2282502746582034
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,160,4096,2688,22,512,32,1,power_law_1.2,2.114186248779297
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,384,4096,2688,22,512,32,1,power_law_1.2,2.231596221923828
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,512,4096,2688,22,512,32,1,power_law_1.2,2.4851654052734373
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1024,4096,2688,22,512,32,1,power_law_1.2,3.239179382324219
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,768,4096,2688,22,512,32,1,power_law_1.2,4.180612487792969
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,1536,4096,2688,22,512,32,1,power_law_1.2,5.179566650390624
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,2048,4096,2688,22,512,32,1,power_law_1.2,4.274851989746094
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,3072,4096,2688,22,512,32,1,power_law_1.2,7.27978515625
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,4096,4096,2688,22,512,32,1,power_law_1.2,8.738345336914062
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,6144,4096,2688,22,512,32,1,power_law_1.2,8.742728881835937
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,float16,20480,4096,2688,22,512,32,1,power_law_1.2,10.455548095703126
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,8192,4096,2688,22,512,32,1,power_law_1.2,14.351644287109375
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,12288,4096,2688,22,512,32,1,power_law_1.2,20.366865234375002
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,16384,4096,2688,22,512,32,1,power_law_1.2,20.84707763671875
SGLang,0.5.9,NVIDIA GB300,moe,sglang_fused_moe_triton,fp8_block,20480,4096,2688,22,512,32,1,power_law_1.2,32.26100830078125
